]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
jbd2: speedup jbd2_journal_dirty_metadata()
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
ac27a0ec 28#include <linux/buffer_head.h>
a5694255 29#include <linux/exportfs.h>
ac27a0ec
DK		 30#include <linux/vfs.h>
31#include <linux/random.h>
32#include <linux/mount.h>
33#include <linux/namei.h>
34#include <linux/quotaops.h>
35#include <linux/seq_file.h>
9f6200bb 36#include <linux/proc_fs.h>
3197ebdb 37#include <linux/ctype.h>
1330593e 38#include <linux/log2.h>
717d50e4 39#include <linux/crc16.h>
7abc52c2 40#include <linux/cleancache.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451 46#include "ext4.h"
4a092d73 47#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
1f109d5a 56static struct proc_dir_entry *ext4_proc_root;
3197ebdb 57static struct kset *ext4_kset;
0b75a840
LC		 58static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
60static struct ext4_features *ext4_feat;
9c191f70 61static int ext4_mballoc_ready;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
2035e776
TT		 78static inline int ext2_feature_set_ok(struct super_block *sb);
79static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 80static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 81static void ext4_destroy_lazyinit_thread(void);
82static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 83static void ext4_clear_request_list(void);
27dd4385 84static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 85
2035e776
TT		 86#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93};
7f78e035 94MODULE_ALIAS_FS("ext2");
fa7614dd 95MODULE_ALIAS("ext2");
2035e776
TT		 96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
ba69f9ab
JK		 102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
152a0836 106 .mount = ext4_mount,
ba69f9ab
JK		 107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
7f78e035 110MODULE_ALIAS_FS("ext3");
fa7614dd 111MODULE_ALIAS("ext3");
ba69f9ab
JK		 112#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113#else
114#define IS_EXT3_SB(sb) (0)
115#endif
bd81d8ee 116
d25425f8
DW		 117static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
119{
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return 1;
123
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125}
126
a9c47317
DW		 127static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
129{
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
132 __u32 csum;
133
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135
136 return cpu_to_le32(csum);
137}
138
c197855e
SH		 139static int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
a9c47317 141{
9aa5d32b 142 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 143 return 1;
144
145 return es->s_checksum == ext4_superblock_csum(sb, es);
146}
147
06db49e6 148void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 149{
06db49e6
TT		 150 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
151
9aa5d32b 152 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 153 return;
154
155 es->s_checksum = ext4_superblock_csum(sb, es);
156}
157
9933fc0a
TT		 158void *ext4_kvmalloc(size_t size, gfp_t flags)
159{
160 void *ret;
161
8be04b93 162 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 163 if (!ret)
164 ret = __vmalloc(size, flags, PAGE_KERNEL);
165 return ret;
166}
167
168void *ext4_kvzalloc(size_t size, gfp_t flags)
169{
170 void *ret;
171
8be04b93 172 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 173 if (!ret)
174 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
175 return ret;
176}
177
8fadc143
AR		 178ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
bd81d8ee 180{
3a14589c 181 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 183 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 184}
185
8fadc143
AR		 186ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
bd81d8ee 188{
5272f837 189 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 192}
193
8fadc143
AR		 194ext4_fsblk_t ext4_inode_table(struct super_block *sb,
195 struct ext4_group_desc *bg)
bd81d8ee 196{
5272f837 197 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 200}
201
021b65bb
TT		 202__u32 ext4_free_group_clusters(struct super_block *sb,
203 struct ext4_group_desc *bg)
560671a0
AK		 204{
205 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 208}
209
210__u32 ext4_free_inodes_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
212{
213 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 215 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 216}
217
218__u32 ext4_used_dirs_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 223 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 224}
225
226__u32 ext4_itable_unused_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_itable_unused_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 231 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 232}
233
8fadc143
AR		 234void ext4_block_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 236{
3a14589c 237 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 240}
241
8fadc143
AR		 242void ext4_inode_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 244{
5272f837 245 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 248}
249
8fadc143
AR		 250void ext4_inode_table_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 252{
5272f837 253 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 256}
257
021b65bb
TT		 258void ext4_free_group_clusters_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 260{
261 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
264}
265
266void ext4_free_inodes_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268{
269 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_used_dirs_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_itable_unused_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
288}
289
d3d1faf6 290
1c13d5c0
TT		 291static void __save_error_info(struct super_block *sb, const char *func,
292 unsigned int line)
293{
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1b46617b
TT		 297 if (bdev_read_only(sb->s_bdev))
298 return;
1c13d5c0
TT		 299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
300 es->s_last_error_time = cpu_to_le32(get_seconds());
301 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
302 es->s_last_error_line = cpu_to_le32(line);
303 if (!es->s_first_error_time) {
304 es->s_first_error_time = es->s_last_error_time;
305 strncpy(es->s_first_error_func, func,
306 sizeof(es->s_first_error_func));
307 es->s_first_error_line = cpu_to_le32(line);
308 es->s_first_error_ino = es->s_last_error_ino;
309 es->s_first_error_block = es->s_last_error_block;
310 }
66e61a9e
TT		 311 /*
312 * Start the daily error reporting function if it hasn't been
313 * started already
314 */
315 if (!es->s_error_count)
316 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 317 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 318}
319
320static void save_error_info(struct super_block *sb, const char *func,
321 unsigned int line)
322{
323 __save_error_info(sb, func, line);
324 ext4_commit_super(sb, 1);
325}
326
18aadd47
BJ		 327static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
328{
329 struct super_block *sb = journal->j_private;
330 struct ext4_sb_info *sbi = EXT4_SB(sb);
331 int error = is_journal_aborted(journal);
5d3ee208 332 struct ext4_journal_cb_entry *jce;
18aadd47 333
5d3ee208 334 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 335 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 336 while (!list_empty(&txn->t_private_list)) {
337 jce = list_entry(txn->t_private_list.next,
338 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 339 list_del_init(&jce->jce_list);
340 spin_unlock(&sbi->s_md_lock);
341 jce->jce_func(sb, jce, error);
342 spin_lock(&sbi->s_md_lock);
343 }
344 spin_unlock(&sbi->s_md_lock);
345}
1c13d5c0 346
ac27a0ec
DK		 347/* Deal with the reporting of failure conditions on a filesystem such as
348 * inconsistencies detected or read IO failures.
349 *
350 * On ext2, we can store the error state of the filesystem in the
617ba13b 351 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 352 * write ordering constraints on the superblock which prevent us from
353 * writing it out straight away; and given that the journal is about to
354 * be aborted, we can't rely on the current, or future, transactions to
355 * write out the superblock safely.
356 *
dab291af 357 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 358 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 359 * that error until we've noted it down and cleared it.
360 */
361
617ba13b 362static void ext4_handle_error(struct super_block *sb)
ac27a0ec 363{
ac27a0ec
DK		 364 if (sb->s_flags & MS_RDONLY)
365 return;
366
2b2d6d01 367 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 368 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 369
4ab2f15b 370 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 371 if (journal)
dab291af 372 jbd2_journal_abort(journal, -EIO);
ac27a0ec 373 }
2b2d6d01 374 if (test_opt(sb, ERRORS_RO)) {
b31e1552 375 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 376 /*
377 * Make sure updated value of ->s_mount_flags will be visible
378 * before ->s_flags update
379 */
380 smp_wmb();
ac27a0ec
DK		 381 sb->s_flags |= MS_RDONLY;
382 }
ac27a0ec 383 if (test_opt(sb, ERRORS_PANIC))
617ba13b 384 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 385 sb->s_id);
386}
387
efbed4dc
TT		 388#define ext4_error_ratelimit(sb) \
389 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
390 "EXT4-fs error")
391
12062ddd 392void __ext4_error(struct super_block *sb, const char *function,
c398eda0 393 unsigned int line, const char *fmt, ...)
ac27a0ec 394{
0ff2ea7d 395 struct va_format vaf;
ac27a0ec
DK		 396 va_list args;
397
efbed4dc
TT		 398 if (ext4_error_ratelimit(sb)) {
399 va_start(args, fmt);
400 vaf.fmt = fmt;
401 vaf.va = &args;
402 printk(KERN_CRIT
403 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
404 sb->s_id, function, line, current->comm, &vaf);
405 va_end(args);
406 }
f3fc0210 407 save_error_info(sb, function, line);
617ba13b 408 ext4_handle_error(sb);
ac27a0ec
DK		 409}
410
e7c96e8e
JP		 411void __ext4_error_inode(struct inode *inode, const char *function,
412 unsigned int line, ext4_fsblk_t block,
413 const char *fmt, ...)
273df556
FM		 414{
415 va_list args;
f7c21177 416 struct va_format vaf;
1c13d5c0 417 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 418
1c13d5c0
TT		 419 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
420 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 421 if (ext4_error_ratelimit(inode->i_sb)) {
422 va_start(args, fmt);
423 vaf.fmt = fmt;
424 vaf.va = &args;
425 if (block)
426 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
427 "inode #%lu: block %llu: comm %s: %pV\n",
428 inode->i_sb->s_id, function, line, inode->i_ino,
429 block, current->comm, &vaf);
430 else
431 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
432 "inode #%lu: comm %s: %pV\n",
433 inode->i_sb->s_id, function, line, inode->i_ino,
434 current->comm, &vaf);
435 va_end(args);
436 }
1c13d5c0 437 save_error_info(inode->i_sb, function, line);
273df556
FM		 438 ext4_handle_error(inode->i_sb);
439}
440
e7c96e8e
JP		 441void __ext4_error_file(struct file *file, const char *function,
442 unsigned int line, ext4_fsblk_t block,
443 const char *fmt, ...)
273df556
FM		 444{
445 va_list args;
f7c21177 446 struct va_format vaf;
1c13d5c0 447 struct ext4_super_block *es;
496ad9aa 448 struct inode *inode = file_inode(file);
273df556
FM		 449 char pathname[80], *path;
450
1c13d5c0
TT		 451 es = EXT4_SB(inode->i_sb)->s_es;
452 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc
TT		 453 if (ext4_error_ratelimit(inode->i_sb)) {
454 path = d_path(&(file->f_path), pathname, sizeof(pathname));
455 if (IS_ERR(path))
456 path = "(unknown)";
457 va_start(args, fmt);
458 vaf.fmt = fmt;
459 vaf.va = &args;
460 if (block)
461 printk(KERN_CRIT
462 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
463 "block %llu: comm %s: path %s: %pV\n",
464 inode->i_sb->s_id, function, line, inode->i_ino,
465 block, current->comm, path, &vaf);
466 else
467 printk(KERN_CRIT
468 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
469 "comm %s: path %s: %pV\n",
470 inode->i_sb->s_id, function, line, inode->i_ino,
471 current->comm, path, &vaf);
472 va_end(args);
473 }
1c13d5c0 474 save_error_info(inode->i_sb, function, line);
273df556
FM		 475 ext4_handle_error(inode->i_sb);
476}
477
722887dd
TT		 478const char *ext4_decode_error(struct super_block *sb, int errno,
479 char nbuf[16])
ac27a0ec
DK		 480{
481 char *errstr = NULL;
482
483 switch (errno) {
484 case -EIO:
485 errstr = "IO failure";
486 break;
487 case -ENOMEM:
488 errstr = "Out of memory";
489 break;
490 case -EROFS:
78f1ddbb
TT		 491 if (!sb || (EXT4_SB(sb)->s_journal &&
492 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 493 errstr = "Journal has aborted";
494 else
495 errstr = "Readonly filesystem";
496 break;
497 default:
498 /* If the caller passed in an extra buffer for unknown
499 * errors, textualise them now. Else we just return
500 * NULL. */
501 if (nbuf) {
502 /* Check for truncated error codes... */
503 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
504 errstr = nbuf;
505 }
506 break;
507 }
508
509 return errstr;
510}
511
617ba13b 512/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 513 * automatically and invokes the appropriate error response. */
514
c398eda0
TT		 515void __ext4_std_error(struct super_block *sb, const char *function,
516 unsigned int line, int errno)
ac27a0ec
DK		 517{
518 char nbuf[16];
519 const char *errstr;
520
521 /* Special case: if the error is EROFS, and we're not already
522 * inside a transaction, then there's really no point in logging
523 * an error. */
524 if (errno == -EROFS && journal_current_handle() == NULL &&
525 (sb->s_flags & MS_RDONLY))
526 return;
527
efbed4dc
TT		 528 if (ext4_error_ratelimit(sb)) {
529 errstr = ext4_decode_error(sb, errno, nbuf);
530 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
531 sb->s_id, function, line, errstr);
532 }
ac27a0ec 533
efbed4dc 534 save_error_info(sb, function, line);
617ba13b 535 ext4_handle_error(sb);
ac27a0ec
DK		 536}
537
538/*
617ba13b 539 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 540 * abort function may be used to deal with unrecoverable failures such
541 * as journal IO errors or ENOMEM at a critical moment in log management.
542 *
543 * We unconditionally force the filesystem into an ABORT|READONLY state,
544 * unless the error response on the fs has been set to panic in which
545 * case we take the easy way out and panic immediately.
546 */
547
c67d859e 548void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 549 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 550{
551 va_list args;
552
1c13d5c0 553 save_error_info(sb, function, line);
ac27a0ec 554 va_start(args, fmt);
c398eda0
TT		 555 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
556 function, line);
ac27a0ec
DK		 557 vprintk(fmt, args);
558 printk("\n");
559 va_end(args);
560
1c13d5c0
TT		 561 if ((sb->s_flags & MS_RDONLY) == 0) {
562 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 563 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 564 /*
565 * Make sure updated value of ->s_mount_flags will be visible
566 * before ->s_flags update
567 */
568 smp_wmb();
569 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 570 if (EXT4_SB(sb)->s_journal)
571 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
572 save_error_info(sb, function, line);
573 }
ac27a0ec 574 if (test_opt(sb, ERRORS_PANIC))
617ba13b 575 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 576}
577
e7c96e8e
JP		 578void __ext4_msg(struct super_block *sb,
579 const char *prefix, const char *fmt, ...)
b31e1552 580{
0ff2ea7d 581 struct va_format vaf;
b31e1552
ES		 582 va_list args;
583
efbed4dc
TT		 584 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
585 return;
586
b31e1552 587 va_start(args, fmt);
0ff2ea7d
JP		 588 vaf.fmt = fmt;
589 vaf.va = &args;
590 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 591 va_end(args);
592}
593
b03a2f7e
AD		 594#define ext4_warning_ratelimit(sb) \
595 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
596 "EXT4-fs warning")
597
12062ddd 598void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 599 unsigned int line, const char *fmt, ...)
ac27a0ec 600{
0ff2ea7d 601 struct va_format vaf;
ac27a0ec
DK		 602 va_list args;
603
b03a2f7e 604 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 605 return;
606
ac27a0ec 607 va_start(args, fmt);
0ff2ea7d
JP		 608 vaf.fmt = fmt;
609 vaf.va = &args;
610 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
611 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 612 va_end(args);
613}
614
b03a2f7e
AD		 615void __ext4_warning_inode(const struct inode *inode, const char *function,
616 unsigned int line, const char *fmt, ...)
617{
618 struct va_format vaf;
619 va_list args;
620
621 if (!ext4_warning_ratelimit(inode->i_sb))
622 return;
623
624 va_start(args, fmt);
625 vaf.fmt = fmt;
626 vaf.va = &args;
627 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
628 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
629 function, line, inode->i_ino, current->comm, &vaf);
630 va_end(args);
631}
632
e29136f8
TT		 633void __ext4_grp_locked_error(const char *function, unsigned int line,
634 struct super_block *sb, ext4_group_t grp,
635 unsigned long ino, ext4_fsblk_t block,
636 const char *fmt, ...)
5d1b1b3f
AK		 637__releases(bitlock)
638__acquires(bitlock)
639{
0ff2ea7d 640 struct va_format vaf;
5d1b1b3f
AK		 641 va_list args;
642 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
643
1c13d5c0
TT		 644 es->s_last_error_ino = cpu_to_le32(ino);
645 es->s_last_error_block = cpu_to_le64(block);
646 __save_error_info(sb, function, line);
0ff2ea7d 647
efbed4dc
TT		 648 if (ext4_error_ratelimit(sb)) {
649 va_start(args, fmt);
650 vaf.fmt = fmt;
651 vaf.va = &args;
652 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
653 sb->s_id, function, line, grp);
654 if (ino)
655 printk(KERN_CONT "inode %lu: ", ino);
656 if (block)
657 printk(KERN_CONT "block %llu:",
658 (unsigned long long) block);
659 printk(KERN_CONT "%pV\n", &vaf);
660 va_end(args);
661 }
5d1b1b3f
AK		 662
663 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 664 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 665 return;
666 }
1c13d5c0 667
5d1b1b3f
AK		 668 ext4_unlock_group(sb, grp);
669 ext4_handle_error(sb);
670 /*
671 * We only get here in the ERRORS_RO case; relocking the group
672 * may be dangerous, but nothing bad will happen since the
673 * filesystem will have already been marked read/only and the
674 * journal has been aborted. We return 1 as a hint to callers
675 * who might what to use the return value from
25985edc 676 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 677 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
678 * aggressively from the ext4 function in question, with a
679 * more appropriate error code.
680 */
681 ext4_lock_group(sb, grp);
682 return;
683}
684
617ba13b 685void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 686{
617ba13b 687 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 688
617ba13b 689 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 690 return;
691
12062ddd 692 ext4_warning(sb,
ac27a0ec
DK		 693 "updating to rev %d because of new feature flag, "
694 "running e2fsck is recommended",
617ba13b 695 EXT4_DYNAMIC_REV);
ac27a0ec 696
617ba13b
MC		 697 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
698 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
699 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 700 /* leave es->s_feature_*compat flags alone */
701 /* es->s_uuid will be set by e2fsck if empty */
702
703 /*
704 * The rest of the superblock fields should be zero, and if not it
705 * means they are likely already in use, so leave them alone. We
706 * can leave it up to e2fsck to clean up any inconsistencies there.
707 */
708}
709
710/*
711 * Open the external journal device
712 */
b31e1552 713static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 714{
715 struct block_device *bdev;
716 char b[BDEVNAME_SIZE];
717
d4d77629 718 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 719 if (IS_ERR(bdev))
720 goto fail;
721 return bdev;
722
723fail:
b31e1552 724 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 725 __bdevname(dev, b), PTR_ERR(bdev));
726 return NULL;
727}
728
729/*
730 * Release the journal device
731 */
4385bab1 732static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 733{
4385bab1 734 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 735}
736
4385bab1 737static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 738{
739 struct block_device *bdev;
ac27a0ec
DK		 740 bdev = sbi->journal_bdev;
741 if (bdev) {
4385bab1 742 ext4_blkdev_put(bdev);
ac27a0ec
DK		 743 sbi->journal_bdev = NULL;
744 }
ac27a0ec
DK		 745}
746
747static inline struct inode *orphan_list_entry(struct list_head *l)
748{
617ba13b 749 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 750}
751
617ba13b 752static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 753{
754 struct list_head *l;
755
b31e1552
ES		 756 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
757 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 758
759 printk(KERN_ERR "sb_info orphan list:\n");
760 list_for_each(l, &sbi->s_orphan) {
761 struct inode *inode = orphan_list_entry(l);
762 printk(KERN_ERR " "
763 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
764 inode->i_sb->s_id, inode->i_ino, inode,
765 inode->i_mode, inode->i_nlink,
766 NEXT_ORPHAN(inode));
767 }
768}
769
2b2d6d01 770static void ext4_put_super(struct super_block *sb)
ac27a0ec 771{
617ba13b
MC		 772 struct ext4_sb_info *sbi = EXT4_SB(sb);
773 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 774 int i, err;
ac27a0ec 775
857ac889 776 ext4_unregister_li_request(sb);
e0ccfd95
CH		 777 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
778
2e8fa54e 779 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 780 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 781
0390131b
FM		 782 if (sbi->s_journal) {
783 err = jbd2_journal_destroy(sbi->s_journal);
784 sbi->s_journal = NULL;
785 if (err < 0)
c67d859e 786 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 787 }
d4edac31 788
d3922a77 789 ext4_es_unregister_shrinker(sbi);
9105bb14 790 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 791 ext4_release_system_zone(sb);
792 ext4_mb_release(sb);
793 ext4_ext_release(sb);
794 ext4_xattr_put_super(sb);
795
ac27a0ec 796 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 797 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 798 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 799 }
58c5873a 800 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 801 ext4_commit_super(sb, 1);
802
240799cd 803 if (sbi->s_proc) {
66acdcf4 804 remove_proc_entry("options", sbi->s_proc);
9f6200bb 805 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 806 }
3197ebdb 807 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 808
809 for (i = 0; i < sbi->s_gdb_count; i++)
810 brelse(sbi->s_group_desc[i]);
b93b41d4
AV		 811 kvfree(sbi->s_group_desc);
812 kvfree(sbi->s_flex_groups);
57042651 813 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 814 percpu_counter_destroy(&sbi->s_freeinodes_counter);
815 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 816 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 817 brelse(sbi->s_sbh);
818#ifdef CONFIG_QUOTA
a2d4a646 819 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 820 kfree(sbi->s_qf_names[i]);
821#endif
822
823 /* Debugging code just in case the in-memory inode orphan list
824 * isn't empty. The on-disk one can be non-empty if we've
825 * detected an error and taken the fs readonly, but the
826 * in-memory list had better be clean by this point. */
827 if (!list_empty(&sbi->s_orphan))
828 dump_orphan_list(sb, sbi);
829 J_ASSERT(list_empty(&sbi->s_orphan));
830
f98393a6 831 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 832 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
833 /*
834 * Invalidate the journal device's buffers. We don't want them
835 * floating about in memory - the physical journal device may
836 * hotswapped, and it breaks the `ro-after' testing code.
837 */
838 sync_blockdev(sbi->journal_bdev);
f98393a6 839 invalidate_bdev(sbi->journal_bdev);
617ba13b 840 ext4_blkdev_remove(sbi);
ac27a0ec 841 }
9c191f70
M		 842 if (sbi->s_mb_cache) {
843 ext4_xattr_destroy_cache(sbi->s_mb_cache);
844 sbi->s_mb_cache = NULL;
845 }
c5e06d10
JL		 846 if (sbi->s_mmp_tsk)
847 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 848 sb->s_fs_info = NULL;
3197ebdb
TT		 849 /*
850 * Now that we are completely done shutting down the
851 * superblock, we need to actually destroy the kobject.
852 */
3197ebdb
TT		 853 kobject_put(&sbi->s_kobj);
854 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 855 if (sbi->s_chksum_driver)
856 crypto_free_shash(sbi->s_chksum_driver);
705895b6 857 kfree(sbi->s_blockgroup_lock);
ac27a0ec 858 kfree(sbi);
ac27a0ec
DK		 859}
860
e18b890b 861static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 862
863/*
864 * Called inside transaction, so use GFP_NOFS
865 */
617ba13b 866static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 867{
617ba13b 868 struct ext4_inode_info *ei;
ac27a0ec 869
e6b4f8da 870 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 871 if (!ei)
872 return NULL;
0b8e58a1 873
ac27a0ec 874 ei->vfs_inode.i_version = 1;
202ee5df 875 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 876 INIT_LIST_HEAD(&ei->i_prealloc_list);
877 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 878 ext4_es_init_tree(&ei->i_es_tree);
879 rwlock_init(&ei->i_es_lock);
edaa53ca 880 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 881 ei->i_es_all_nr = 0;
edaa53ca 882 ei->i_es_shk_nr = 0;
dd475925 883 ei->i_es_shrink_lblk = 0;
d2a17637
MC		 884 ei->i_reserved_data_blocks = 0;
885 ei->i_reserved_meta_blocks = 0;
886 ei->i_allocated_meta_blocks = 0;
9d0be502 887 ei->i_da_metadata_calc_len = 0;
7e731bc9 888 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 889 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 890#ifdef CONFIG_QUOTA
891 ei->i_reserved_quota = 0;
96c7e0d9 892 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 893#endif
8aefcd55 894 ei->jinode = NULL;
2e8fa54e 895 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 896 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 897 ei->i_sync_tid = 0;
898 ei->i_datasync_tid = 0;
f7ad6d2e 899 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 900 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 901 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
b30ab0e0 902#ifdef CONFIG_EXT4_FS_ENCRYPTION
b7236e21 903 ei->i_crypt_info = NULL;
b30ab0e0 904#endif
ac27a0ec
DK		 905 return &ei->vfs_inode;
906}
907
7ff9c073
TT		 908static int ext4_drop_inode(struct inode *inode)
909{
910 int drop = generic_drop_inode(inode);
911
912 trace_ext4_drop_inode(inode, drop);
913 return drop;
914}
915
fa0d7e3d
NP		 916static void ext4_i_callback(struct rcu_head *head)
917{
918 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 919 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
920}
921
617ba13b 922static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 923{
9f7dd93d 924 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 925 ext4_msg(inode->i_sb, KERN_ERR,
926 "Inode %lu (%p): orphan list check failed!",
927 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 928 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
929 EXT4_I(inode), sizeof(struct ext4_inode_info),
930 true);
931 dump_stack();
932 }
fa0d7e3d 933 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 934}
935
51cc5068 936static void init_once(void *foo)
ac27a0ec 937{
617ba13b 938 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 939
a35afb83 940 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 941 init_rwsem(&ei->xattr_sem);
0e855ac8 942 init_rwsem(&ei->i_data_sem);
a35afb83 943 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 944}
945
e67bc2b3 946static int __init init_inodecache(void)
ac27a0ec 947{
617ba13b
MC		 948 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
949 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 950 0, (SLAB_RECLAIM_ACCOUNT|
951 SLAB_MEM_SPREAD),
20c2df83 952 init_once);
617ba13b 953 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 954 return -ENOMEM;
955 return 0;
956}
957
958static void destroy_inodecache(void)
959{
8c0a8537
KS		 960 /*
961 * Make sure all delayed rcu free inodes are flushed before we
962 * destroy cache.
963 */
964 rcu_barrier();
617ba13b 965 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 966}
967
0930fcc1 968void ext4_clear_inode(struct inode *inode)
ac27a0ec 969{
0930fcc1 970 invalidate_inode_buffers(inode);
dbd5768f 971 clear_inode(inode);
9f754758 972 dquot_drop(inode);
c2ea3fde 973 ext4_discard_preallocations(inode);
51865fda 974 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 975 if (EXT4_I(inode)->jinode) {
976 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
977 EXT4_I(inode)->jinode);
978 jbd2_free_inode(EXT4_I(inode)->jinode);
979 EXT4_I(inode)->jinode = NULL;
980 }
b7236e21
TT		 981#ifdef CONFIG_EXT4_FS_ENCRYPTION
982 if (EXT4_I(inode)->i_crypt_info)
c936e1ec 983 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
b7236e21 984#endif
ac27a0ec
DK		 985}
986
1b961ac0 987static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 988 u64 ino, u32 generation)
ac27a0ec 989{
ac27a0ec 990 struct inode *inode;
ac27a0ec 991
617ba13b 992 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 993 return ERR_PTR(-ESTALE);
617ba13b 994 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 995 return ERR_PTR(-ESTALE);
996
997 /* iget isn't really right if the inode is currently unallocated!!
998 *
617ba13b 999 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1000 * deleted, so we should be safe.
1001 *
1002 * Currently we don't know the generation for parent directory, so
1003 * a generation of 0 means "accept any"
1004 */
f4bb2981 1005 inode = ext4_iget_normal(sb, ino);
1d1fe1ee
DH		 1006 if (IS_ERR(inode))
1007 return ERR_CAST(inode);
1008 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1009 iput(inode);
1010 return ERR_PTR(-ESTALE);
1011 }
1b961ac0
CH		 1012
1013 return inode;
1014}
1015
1016static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1017 int fh_len, int fh_type)
1b961ac0
CH		 1018{
1019 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1020 ext4_nfs_get_inode);
1021}
1022
1023static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1024 int fh_len, int fh_type)
1b961ac0
CH		 1025{
1026 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1027 ext4_nfs_get_inode);
ac27a0ec
DK		 1028}
1029
c39a7f84
TO		 1030/*
1031 * Try to release metadata pages (indirect blocks, directories) which are
1032 * mapped via the block device. Since these pages could have journal heads
1033 * which would prevent try_to_free_buffers() from freeing them, we must use
1034 * jbd2 layer's try_to_free_buffers() function to release them.
1035 */
0b8e58a1
AD		 1036static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1037 gfp_t wait)
c39a7f84
TO		 1038{
1039 journal_t *journal = EXT4_SB(sb)->s_journal;
1040
1041 WARN_ON(PageChecked(page));
1042 if (!page_has_buffers(page))
1043 return 0;
1044 if (journal)
1045 return jbd2_journal_try_to_free_buffers(journal, page,
1046 wait & ~__GFP_WAIT);
1047 return try_to_free_buffers(page);
1048}
1049
ac27a0ec 1050#ifdef CONFIG_QUOTA
af5bc92d 1051#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1052#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1053
617ba13b
MC		 1054static int ext4_write_dquot(struct dquot *dquot);
1055static int ext4_acquire_dquot(struct dquot *dquot);
1056static int ext4_release_dquot(struct dquot *dquot);
1057static int ext4_mark_dquot_dirty(struct dquot *dquot);
1058static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1059static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1060 struct path *path);
ca0e05e4 1061static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1062static int ext4_quota_on_mount(struct super_block *sb, int type);
1063static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1064 size_t len, loff_t off);
617ba13b 1065static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1066 const char *data, size_t len, loff_t off);
7c319d32
AK		 1067static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1068 unsigned int flags);
1069static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1070
96c7e0d9
JK		 1071static struct dquot **ext4_get_dquots(struct inode *inode)
1072{
1073 return EXT4_I(inode)->i_dquot;
1074}
1075
61e225dc 1076static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1077 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1078 .write_dquot = ext4_write_dquot,
1079 .acquire_dquot = ext4_acquire_dquot,
1080 .release_dquot = ext4_release_dquot,
1081 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1082 .write_info = ext4_write_info,
1083 .alloc_dquot = dquot_alloc,
1084 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1085};
1086
0d54b217 1087static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1088 .quota_on = ext4_quota_on,
ca0e05e4 1089 .quota_off = ext4_quota_off,
287a8095 1090 .quota_sync = dquot_quota_sync,
0a240339 1091 .get_state = dquot_get_state,
287a8095
CH		 1092 .set_info = dquot_set_dqinfo,
1093 .get_dqblk = dquot_get_dqblk,
1094 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1095};
1096#endif
1097
ee9b6d61 1098static const struct super_operations ext4_sops = {
617ba13b
MC		 1099 .alloc_inode = ext4_alloc_inode,
1100 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1101 .write_inode = ext4_write_inode,
1102 .dirty_inode = ext4_dirty_inode,
7ff9c073 1103 .drop_inode = ext4_drop_inode,
0930fcc1 1104 .evict_inode = ext4_evict_inode,
617ba13b 1105 .put_super = ext4_put_super,
617ba13b 1106 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1107 .freeze_fs = ext4_freeze,
1108 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1109 .statfs = ext4_statfs,
1110 .remount_fs = ext4_remount,
617ba13b 1111 .show_options = ext4_show_options,
ac27a0ec 1112#ifdef CONFIG_QUOTA
617ba13b
MC		 1113 .quota_read = ext4_quota_read,
1114 .quota_write = ext4_quota_write,
96c7e0d9 1115 .get_dquots = ext4_get_dquots,
ac27a0ec 1116#endif
c39a7f84 1117 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1118};
1119
39655164 1120static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1121 .fh_to_dentry = ext4_fh_to_dentry,
1122 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1123 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1124};
1125
1126enum {
1127 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1128 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1129 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1130 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1131 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1132 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1133 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1134 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1135 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
ac27a0ec 1136 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1137 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1138 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
923ae0ff 1139 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1449032b 1140 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
a26f4992 1141 Opt_lazytime, Opt_nolazytime,
1449032b 1142 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1143 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1144 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1145 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
c6d3d56d 1146 Opt_max_dir_size_kb, Opt_nojournal_checksum,
ac27a0ec
DK		 1147};
1148
a447c093 1149static const match_table_t tokens = {
ac27a0ec
DK		 1150 {Opt_bsd_df, "bsddf"},
1151 {Opt_minix_df, "minixdf"},
1152 {Opt_grpid, "grpid"},
1153 {Opt_grpid, "bsdgroups"},
1154 {Opt_nogrpid, "nogrpid"},
1155 {Opt_nogrpid, "sysvgroups"},
1156 {Opt_resgid, "resgid=%u"},
1157 {Opt_resuid, "resuid=%u"},
1158 {Opt_sb, "sb=%u"},
1159 {Opt_err_cont, "errors=continue"},
1160 {Opt_err_panic, "errors=panic"},
1161 {Opt_err_ro, "errors=remount-ro"},
1162 {Opt_nouid32, "nouid32"},
ac27a0ec 1163 {Opt_debug, "debug"},
72578c33
TT		 1164 {Opt_removed, "oldalloc"},
1165 {Opt_removed, "orlov"},
ac27a0ec
DK		 1166 {Opt_user_xattr, "user_xattr"},
1167 {Opt_nouser_xattr, "nouser_xattr"},
1168 {Opt_acl, "acl"},
1169 {Opt_noacl, "noacl"},
e3bb52ae 1170 {Opt_noload, "norecovery"},
5a916be1 1171 {Opt_noload, "noload"},
72578c33
TT		 1172 {Opt_removed, "nobh"},
1173 {Opt_removed, "bh"},
ac27a0ec 1174 {Opt_commit, "commit=%u"},
30773840
TT		 1175 {Opt_min_batch_time, "min_batch_time=%u"},
1176 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1177 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1178 {Opt_journal_path, "journal_path=%s"},
818d276c 1179 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1180 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1181 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1182 {Opt_abort, "abort"},
1183 {Opt_data_journal, "data=journal"},
1184 {Opt_data_ordered, "data=ordered"},
1185 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1186 {Opt_data_err_abort, "data_err=abort"},
1187 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1188 {Opt_offusrjquota, "usrjquota="},
1189 {Opt_usrjquota, "usrjquota=%s"},
1190 {Opt_offgrpjquota, "grpjquota="},
1191 {Opt_grpjquota, "grpjquota=%s"},
1192 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1193 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1194 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1195 {Opt_grpquota, "grpquota"},
1196 {Opt_noquota, "noquota"},
1197 {Opt_quota, "quota"},
1198 {Opt_usrquota, "usrquota"},
1199 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1200 {Opt_barrier, "barrier"},
1201 {Opt_nobarrier, "nobarrier"},
25ec56b5 1202 {Opt_i_version, "i_version"},
923ae0ff 1203 {Opt_dax, "dax"},
c9de560d 1204 {Opt_stripe, "stripe=%u"},
64769240 1205 {Opt_delalloc, "delalloc"},
a26f4992
TT		 1206 {Opt_lazytime, "lazytime"},
1207 {Opt_nolazytime, "nolazytime"},
dd919b98 1208 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1209 {Opt_removed, "mblk_io_submit"},
1210 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1211 {Opt_block_validity, "block_validity"},
1212 {Opt_noblock_validity, "noblock_validity"},
240799cd 1213 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1214 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1215 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1216 {Opt_auto_da_alloc, "auto_da_alloc"},
1217 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1218 {Opt_dioread_nolock, "dioread_nolock"},
1219 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1220 {Opt_discard, "discard"},
1221 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1222 {Opt_init_itable, "init_itable=%u"},
1223 {Opt_init_itable, "init_itable"},
1224 {Opt_noinit_itable, "noinit_itable"},
df981d03 1225 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
6ddb2447 1226 {Opt_test_dummy_encryption, "test_dummy_encryption"},
c7198b9c
TT		 1227 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1228 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1229 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1230 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1231 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1232 {Opt_err, NULL},
ac27a0ec
DK		 1233};
1234
617ba13b 1235static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1236{
617ba13b 1237 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1238 char *options = (char *) *data;
1239
1240 if (!options || strncmp(options, "sb=", 3) != 0)
1241 return 1; /* Default location */
0b8e58a1 1242
ac27a0ec 1243 options += 3;
0b8e58a1 1244 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1245 sb_block = simple_strtoul(options, &options, 0);
1246 if (*options && *options != ',') {
4776004f 1247 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1248 (char *) *data);
1249 return 1;
1250 }
1251 if (*options == ',')
1252 options++;
1253 *data = (void *) options;
0b8e58a1 1254
ac27a0ec
DK		 1255 return sb_block;
1256}
1257
b3881f74 1258#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1259static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1260 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1261
56c50f11
DM		 1262#ifdef CONFIG_QUOTA
1263static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1264{
1265 struct ext4_sb_info *sbi = EXT4_SB(sb);
1266 char *qname;
03dafb5f 1267 int ret = -1;
56c50f11
DM		 1268
1269 if (sb_any_quota_loaded(sb) &&
1270 !sbi->s_qf_names[qtype]) {
1271 ext4_msg(sb, KERN_ERR,
1272 "Cannot change journaled "
1273 "quota options when quota turned on");
57f73c2c 1274 return -1;
56c50f11 1275 }
262b4662
JK		 1276 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1277 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1278 "when QUOTA feature is enabled");
1279 return -1;
1280 }
56c50f11
DM		 1281 qname = match_strdup(args);
1282 if (!qname) {
1283 ext4_msg(sb, KERN_ERR,
1284 "Not enough memory for storing quotafile name");
57f73c2c 1285 return -1;
56c50f11 1286 }
03dafb5f
CG		 1287 if (sbi->s_qf_names[qtype]) {
1288 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1289 ret = 1;
1290 else
1291 ext4_msg(sb, KERN_ERR,
1292 "%s quota file already specified",
1293 QTYPE2NAME(qtype));
1294 goto errout;
56c50f11 1295 }
03dafb5f 1296 if (strchr(qname, '/')) {
56c50f11
DM		 1297 ext4_msg(sb, KERN_ERR,
1298 "quotafile must be on filesystem root");
03dafb5f 1299 goto errout;
56c50f11 1300 }
03dafb5f 1301 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1302 set_opt(sb, QUOTA);
56c50f11 1303 return 1;
03dafb5f
CG		 1304errout:
1305 kfree(qname);
1306 return ret;
56c50f11
DM		 1307}
1308
1309static int clear_qf_name(struct super_block *sb, int qtype)
1310{
1311
1312 struct ext4_sb_info *sbi = EXT4_SB(sb);
1313
1314 if (sb_any_quota_loaded(sb) &&
1315 sbi->s_qf_names[qtype]) {
1316 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1317 " when quota turned on");
57f73c2c 1318 return -1;
56c50f11 1319 }
03dafb5f 1320 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1321 sbi->s_qf_names[qtype] = NULL;
1322 return 1;
1323}
1324#endif
1325
26092bf5
TT		 1326#define MOPT_SET 0x0001
1327#define MOPT_CLEAR 0x0002
1328#define MOPT_NOSUPPORT 0x0004
1329#define MOPT_EXPLICIT 0x0008
1330#define MOPT_CLEAR_ERR 0x0010
1331#define MOPT_GTE0 0x0020
ac27a0ec 1332#ifdef CONFIG_QUOTA
26092bf5
TT		 1333#define MOPT_Q 0
1334#define MOPT_QFMT 0x0040
1335#else
1336#define MOPT_Q MOPT_NOSUPPORT
1337#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1338#endif
26092bf5 1339#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1340#define MOPT_NO_EXT2 0x0100
1341#define MOPT_NO_EXT3 0x0200
1342#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1343#define MOPT_STRING 0x0400
26092bf5
TT		 1344
1345static const struct mount_opts {
1346 int token;
1347 int mount_opt;
1348 int flags;
1349} ext4_mount_opts[] = {
1350 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1351 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1352 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1353 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1354 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1355 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1356 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1357 MOPT_EXT4_ONLY | MOPT_SET},
1358 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1359 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1360 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1361 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1362 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1363 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1364 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1365 MOPT_EXT4_ONLY | MOPT_CLEAR},
c6d3d56d
DW		 1366 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1367 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1368 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1369 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1370 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1371 EXT4_MOUNT_JOURNAL_CHECKSUM),
1372 MOPT_EXT4_ONLY | MOPT_SET},
1373 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1374 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1375 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1376 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1377 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1378 MOPT_NO_EXT2 | MOPT_SET},
1379 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1380 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1381 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1382 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1383 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1384 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1385 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1386 {Opt_commit, 0, MOPT_GTE0},
1387 {Opt_max_batch_time, 0, MOPT_GTE0},
1388 {Opt_min_batch_time, 0, MOPT_GTE0},
1389 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1390 {Opt_init_itable, 0, MOPT_GTE0},
923ae0ff 1391 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
26092bf5 1392 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1393 {Opt_resuid, 0, MOPT_GTE0},
1394 {Opt_resgid, 0, MOPT_GTE0},
1395 {Opt_journal_dev, 0, MOPT_GTE0},
ad4eec61 1396 {Opt_journal_path, 0, MOPT_STRING},
0efb3b23 1397 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1398 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1399 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1400 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1401 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1402 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1403 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1404#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1405 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1406 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1407#else
26092bf5
TT		 1408 {Opt_acl, 0, MOPT_NOSUPPORT},
1409 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1410#endif
26092bf5
TT		 1411 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1412 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1413 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1414 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1415 MOPT_SET | MOPT_Q},
1416 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1417 MOPT_SET | MOPT_Q},
1418 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1419 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1420 {Opt_usrjquota, 0, MOPT_Q},
1421 {Opt_grpjquota, 0, MOPT_Q},
1422 {Opt_offusrjquota, 0, MOPT_Q},
1423 {Opt_offgrpjquota, 0, MOPT_Q},
1424 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1425 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1426 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1427 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
6ddb2447 1428 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
26092bf5
TT		 1429 {Opt_err, 0, 0}
1430};
1431
1432static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1433 substring_t *args, unsigned long *journal_devnum,
1434 unsigned int *journal_ioprio, int is_remount)
1435{
1436 struct ext4_sb_info *sbi = EXT4_SB(sb);
1437 const struct mount_opts *m;
08cefc7a
EB		 1438 kuid_t uid;
1439 kgid_t gid;
26092bf5
TT		 1440 int arg = 0;
1441
57f73c2c
TT		 1442#ifdef CONFIG_QUOTA
1443 if (token == Opt_usrjquota)
1444 return set_qf_name(sb, USRQUOTA, &args[0]);
1445 else if (token == Opt_grpjquota)
1446 return set_qf_name(sb, GRPQUOTA, &args[0]);
1447 else if (token == Opt_offusrjquota)
1448 return clear_qf_name(sb, USRQUOTA);
1449 else if (token == Opt_offgrpjquota)
1450 return clear_qf_name(sb, GRPQUOTA);
1451#endif
26092bf5 1452 switch (token) {
f7048605
TT		 1453 case Opt_noacl:
1454 case Opt_nouser_xattr:
1455 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1456 break;
26092bf5
TT		 1457 case Opt_sb:
1458 return 1; /* handled by get_sb_block() */
1459 case Opt_removed:
5f3633e3 1460 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1461 return 1;
26092bf5
TT		 1462 case Opt_abort:
1463 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1464 return 1;
1465 case Opt_i_version:
1466 sb->s_flags |= MS_I_VERSION;
1467 return 1;
a26f4992
TT		 1468 case Opt_lazytime:
1469 sb->s_flags |= MS_LAZYTIME;
1470 return 1;
1471 case Opt_nolazytime:
1472 sb->s_flags &= ~MS_LAZYTIME;
1473 return 1;
26092bf5
TT		 1474 }
1475
5f3633e3
JK		 1476 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1477 if (token == m->token)
1478 break;
1479
1480 if (m->token == Opt_err) {
1481 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1482 "or missing value", opt);
1483 return -1;
1484 }
1485
8dc0aa8c
TT		 1486 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1487 ext4_msg(sb, KERN_ERR,
1488 "Mount option \"%s\" incompatible with ext2", opt);
1489 return -1;
1490 }
1491 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1492 ext4_msg(sb, KERN_ERR,
1493 "Mount option \"%s\" incompatible with ext3", opt);
1494 return -1;
1495 }
1496
ad4eec61 1497 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1498 return -1;
1499 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1500 return -1;
1501 if (m->flags & MOPT_EXPLICIT)
1502 set_opt2(sb, EXPLICIT_DELALLOC);
1503 if (m->flags & MOPT_CLEAR_ERR)
1504 clear_opt(sb, ERRORS_MASK);
1505 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1506 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1507 "options when quota turned on");
1508 return -1;
1509 }
1510
1511 if (m->flags & MOPT_NOSUPPORT) {
1512 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1513 } else if (token == Opt_commit) {
1514 if (arg == 0)
1515 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1516 sbi->s_commit_interval = HZ * arg;
1517 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1518 sbi->s_max_batch_time = arg;
1519 } else if (token == Opt_min_batch_time) {
1520 sbi->s_min_batch_time = arg;
1521 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1522 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1523 ext4_msg(sb, KERN_ERR,
1524 "EXT4-fs: inode_readahead_blks must be "
1525 "0 or a power of 2 smaller than 2^31");
26092bf5 1526 return -1;
5f3633e3
JK		 1527 }
1528 sbi->s_inode_readahead_blks = arg;
1529 } else if (token == Opt_init_itable) {
1530 set_opt(sb, INIT_INODE_TABLE);
1531 if (!args->from)
1532 arg = EXT4_DEF_LI_WAIT_MULT;
1533 sbi->s_li_wait_mult = arg;
1534 } else if (token == Opt_max_dir_size_kb) {
1535 sbi->s_max_dir_size_kb = arg;
1536 } else if (token == Opt_stripe) {
1537 sbi->s_stripe = arg;
1538 } else if (token == Opt_resuid) {
1539 uid = make_kuid(current_user_ns(), arg);
1540 if (!uid_valid(uid)) {
1541 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1542 return -1;
1543 }
5f3633e3
JK		 1544 sbi->s_resuid = uid;
1545 } else if (token == Opt_resgid) {
1546 gid = make_kgid(current_user_ns(), arg);
1547 if (!gid_valid(gid)) {
1548 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1549 return -1;
1550 }
1551 sbi->s_resgid = gid;
1552 } else if (token == Opt_journal_dev) {
1553 if (is_remount) {
1554 ext4_msg(sb, KERN_ERR,
1555 "Cannot specify journal on remount");
1556 return -1;
1557 }
1558 *journal_devnum = arg;
ad4eec61
ES		 1559 } else if (token == Opt_journal_path) {
1560 char *journal_path;
1561 struct inode *journal_inode;
1562 struct path path;
1563 int error;
1564
1565 if (is_remount) {
1566 ext4_msg(sb, KERN_ERR,
1567 "Cannot specify journal on remount");
1568 return -1;
1569 }
1570 journal_path = match_strdup(&args[0]);
1571 if (!journal_path) {
1572 ext4_msg(sb, KERN_ERR, "error: could not dup "
1573 "journal device string");
1574 return -1;
1575 }
1576
1577 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1578 if (error) {
1579 ext4_msg(sb, KERN_ERR, "error: could not find "
1580 "journal device path: error %d", error);
1581 kfree(journal_path);
1582 return -1;
1583 }
1584
2b0143b5 1585 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 1586 if (!S_ISBLK(journal_inode->i_mode)) {
1587 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1588 "is not a block device", journal_path);
1589 path_put(&path);
1590 kfree(journal_path);
1591 return -1;
1592 }
1593
1594 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1595 path_put(&path);
1596 kfree(journal_path);
5f3633e3
JK		 1597 } else if (token == Opt_journal_ioprio) {
1598 if (arg > 7) {
1599 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1600 " (must be 0-7)");
1601 return -1;
1602 }
1603 *journal_ioprio =
1604 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447
TT		 1605 } else if (token == Opt_test_dummy_encryption) {
1606#ifdef CONFIG_EXT4_FS_ENCRYPTION
1607 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1608 ext4_msg(sb, KERN_WARNING,
1609 "Test dummy encryption mode enabled");
1610#else
1611 ext4_msg(sb, KERN_WARNING,
1612 "Test dummy encryption mount option ignored");
1613#endif
5f3633e3
JK		 1614 } else if (m->flags & MOPT_DATAJ) {
1615 if (is_remount) {
1616 if (!sbi->s_journal)
1617 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1618 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1619 ext4_msg(sb, KERN_ERR,
26092bf5 1620 "Cannot change data mode on remount");
26092bf5 1621 return -1;
ac27a0ec 1622 }
26092bf5 1623 } else {
5f3633e3
JK		 1624 clear_opt(sb, DATA_FLAGS);
1625 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1626 }
5f3633e3
JK		 1627#ifdef CONFIG_QUOTA
1628 } else if (m->flags & MOPT_QFMT) {
1629 if (sb_any_quota_loaded(sb) &&
1630 sbi->s_jquota_fmt != m->mount_opt) {
1631 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1632 "quota options when quota turned on");
1633 return -1;
1634 }
262b4662
JK		 1635 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1636 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1637 ext4_msg(sb, KERN_ERR,
1638 "Cannot set journaled quota options "
1639 "when QUOTA feature is enabled");
1640 return -1;
1641 }
5f3633e3 1642 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff
RZ		 1643#endif
1644#ifndef CONFIG_FS_DAX
1645 } else if (token == Opt_dax) {
1646 ext4_msg(sb, KERN_INFO, "dax option not supported");
1647 return -1;
5f3633e3
JK		 1648#endif
1649 } else {
1650 if (!args->from)
1651 arg = 1;
1652 if (m->flags & MOPT_CLEAR)
1653 arg = !arg;
1654 else if (unlikely(!(m->flags & MOPT_SET))) {
1655 ext4_msg(sb, KERN_WARNING,
1656 "buggy handling of option %s", opt);
1657 WARN_ON(1);
1658 return -1;
1659 }
1660 if (arg != 0)
1661 sbi->s_mount_opt |= m->mount_opt;
1662 else
1663 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1664 }
5f3633e3 1665 return 1;
26092bf5
TT		 1666}
1667
1668static int parse_options(char *options, struct super_block *sb,
1669 unsigned long *journal_devnum,
1670 unsigned int *journal_ioprio,
1671 int is_remount)
1672{
1673 struct ext4_sb_info *sbi = EXT4_SB(sb);
1674 char *p;
1675 substring_t args[MAX_OPT_ARGS];
1676 int token;
1677
1678 if (!options)
1679 return 1;
1680
1681 while ((p = strsep(&options, ",")) != NULL) {
1682 if (!*p)
1683 continue;
1684 /*
1685 * Initialize args struct so we know whether arg was
1686 * found; some options take optional arguments.
1687 */
caecd0af 1688 args[0].to = args[0].from = NULL;
26092bf5
TT		 1689 token = match_token(p, tokens, args);
1690 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1691 journal_ioprio, is_remount) < 0)
1692 return 0;
ac27a0ec
DK		 1693 }
1694#ifdef CONFIG_QUOTA
262b4662
JK		 1695 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1696 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1697 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1698 "feature is enabled");
1699 return 0;
1700 }
ac27a0ec 1701 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1702 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1703 clear_opt(sb, USRQUOTA);
ac27a0ec 1704
482a7425 1705 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1706 clear_opt(sb, GRPQUOTA);
ac27a0ec 1707
56c50f11 1708 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1709 ext4_msg(sb, KERN_ERR, "old and new quota "
1710 "format mixing");
ac27a0ec
DK		 1711 return 0;
1712 }
1713
1714 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1715 ext4_msg(sb, KERN_ERR, "journaled quota format "
1716 "not specified");
ac27a0ec
DK		 1717 return 0;
1718 }
ac27a0ec
DK		 1719 }
1720#endif
261cb20c
JK		 1721 if (test_opt(sb, DIOREAD_NOLOCK)) {
1722 int blocksize =
1723 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1724
1725 if (blocksize < PAGE_CACHE_SIZE) {
1726 ext4_msg(sb, KERN_ERR, "can't mount with "
1727 "dioread_nolock if block size != PAGE_SIZE");
1728 return 0;
1729 }
1730 }
d4f76107
JK		 1731 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1732 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1733 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1734 "in data=ordered mode");
1735 return 0;
1736 }
ac27a0ec
DK		 1737 return 1;
1738}
1739
2adf6da8
TT		 1740static inline void ext4_show_quota_options(struct seq_file *seq,
1741 struct super_block *sb)
1742{
1743#if defined(CONFIG_QUOTA)
1744 struct ext4_sb_info *sbi = EXT4_SB(sb);
1745
1746 if (sbi->s_jquota_fmt) {
1747 char *fmtname = "";
1748
1749 switch (sbi->s_jquota_fmt) {
1750 case QFMT_VFS_OLD:
1751 fmtname = "vfsold";
1752 break;
1753 case QFMT_VFS_V0:
1754 fmtname = "vfsv0";
1755 break;
1756 case QFMT_VFS_V1:
1757 fmtname = "vfsv1";
1758 break;
1759 }
1760 seq_printf(seq, ",jqfmt=%s", fmtname);
1761 }
1762
1763 if (sbi->s_qf_names[USRQUOTA])
1764 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1765
1766 if (sbi->s_qf_names[GRPQUOTA])
1767 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1768#endif
1769}
1770
5a916be1
TT		 1771static const char *token2str(int token)
1772{
50df9fd5 1773 const struct match_token *t;
5a916be1
TT		 1774
1775 for (t = tokens; t->token != Opt_err; t++)
1776 if (t->token == token && !strchr(t->pattern, '='))
1777 break;
1778 return t->pattern;
1779}
1780
2adf6da8
TT		 1781/*
1782 * Show an option if
1783 * - it's set to a non-default value OR
1784 * - if the per-sb default is different from the global default
1785 */
66acdcf4
TT		 1786static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1787 int nodefs)
2adf6da8 1788{
2adf6da8
TT		 1789 struct ext4_sb_info *sbi = EXT4_SB(sb);
1790 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1791 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1792 const struct mount_opts *m;
66acdcf4 1793 char sep = nodefs ? '\n' : ',';
2adf6da8 1794
66acdcf4
TT		 1795#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1796#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1797
1798 if (sbi->s_sb_block != 1)
5a916be1
TT		 1799 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1800
1801 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1802 int want_set = m->flags & MOPT_SET;
1803 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1804 (m->flags & MOPT_CLEAR_ERR))
1805 continue;
66acdcf4 1806 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1807 continue; /* skip if same as the default */
1808 if ((want_set &&
1809 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1810 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1811 continue; /* select Opt_noFoo vs Opt_Foo */
1812 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1813 }
5a916be1 1814
08cefc7a 1815 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1816 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1817 SEQ_OPTS_PRINT("resuid=%u",
1818 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1819 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1820 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1821 SEQ_OPTS_PRINT("resgid=%u",
1822 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1823 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1824 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1825 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1826 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1827 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1828 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1829 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1830 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1831 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1832 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1833 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1834 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1835 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1836 if (sb->s_flags & MS_I_VERSION)
5a916be1 1837 SEQ_OPTS_PUTS("i_version");
66acdcf4 1838 if (nodefs || sbi->s_stripe)
5a916be1 1839 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1840 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1841 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1842 SEQ_OPTS_PUTS("data=journal");
1843 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1844 SEQ_OPTS_PUTS("data=ordered");
1845 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1846 SEQ_OPTS_PUTS("data=writeback");
1847 }
66acdcf4
TT		 1848 if (nodefs ||
1849 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1850 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1851 sbi->s_inode_readahead_blks);
2adf6da8 1852
66acdcf4
TT		 1853 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1854 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1855 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1856 if (nodefs || sbi->s_max_dir_size_kb)
1857 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1858
1859 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1860 return 0;
1861}
1862
66acdcf4
TT		 1863static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1864{
1865 return _ext4_show_options(seq, root->d_sb, 0);
1866}
1867
1868static int options_seq_show(struct seq_file *seq, void *offset)
1869{
1870 struct super_block *sb = seq->private;
1871 int rc;
1872
1873 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1874 rc = _ext4_show_options(seq, sb, 1);
1875 seq_puts(seq, "\n");
1876 return rc;
1877}
1878
1879static int options_open_fs(struct inode *inode, struct file *file)
1880{
d9dda78b 1881 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1882}
1883
1884static const struct file_operations ext4_seq_options_fops = {
1885 .owner = THIS_MODULE,
1886 .open = options_open_fs,
1887 .read = seq_read,
1888 .llseek = seq_lseek,
1889 .release = single_release,
1890};
1891
617ba13b 1892static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1893 int read_only)
1894{
617ba13b 1895 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1896 int res = 0;
1897
617ba13b 1898 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1899 ext4_msg(sb, KERN_ERR, "revision level too high, "
1900 "forcing read-only mode");
ac27a0ec
DK		 1901 res = MS_RDONLY;
1902 }
1903 if (read_only)
281b5995 1904 goto done;
617ba13b 1905 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1906 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1907 "running e2fsck is recommended");
c8b459f4 1908 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1909 ext4_msg(sb, KERN_WARNING,
1910 "warning: mounting fs with errors, "
1911 "running e2fsck is recommended");
ed3ce80a 1912 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1913 le16_to_cpu(es->s_mnt_count) >=
1914 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1915 ext4_msg(sb, KERN_WARNING,
1916 "warning: maximal mount count reached, "
1917 "running e2fsck is recommended");
ac27a0ec
DK		 1918 else if (le32_to_cpu(es->s_checkinterval) &&
1919 (le32_to_cpu(es->s_lastcheck) +
1920 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1921 ext4_msg(sb, KERN_WARNING,
1922 "warning: checktime reached, "
1923 "running e2fsck is recommended");
0b8e58a1 1924 if (!sbi->s_journal)
0390131b 1925 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1926 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1927 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1928 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1929 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1930 ext4_update_dynamic_rev(sb);
0390131b
FM		 1931 if (sbi->s_journal)
1932 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1933
e2d67052 1934 ext4_commit_super(sb, 1);
281b5995 1935done:
ac27a0ec 1936 if (test_opt(sb, DEBUG))
a9df9a49 1937 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1938 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1939 sb->s_blocksize,
1940 sbi->s_groups_count,
617ba13b
MC		 1941 EXT4_BLOCKS_PER_GROUP(sb),
1942 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1943 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1944
7abc52c2 1945 cleancache_init_fs(sb);
ac27a0ec
DK		 1946 return res;
1947}
1948
117fff10
TT		 1949int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1950{
1951 struct ext4_sb_info *sbi = EXT4_SB(sb);
1952 struct flex_groups *new_groups;
1953 int size;
1954
1955 if (!sbi->s_log_groups_per_flex)
1956 return 0;
1957
1958 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1959 if (size <= sbi->s_flex_groups_allocated)
1960 return 0;
1961
1962 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1963 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1964 if (!new_groups) {
1965 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1966 size / (int) sizeof(struct flex_groups));
1967 return -ENOMEM;
1968 }
1969
1970 if (sbi->s_flex_groups) {
1971 memcpy(new_groups, sbi->s_flex_groups,
1972 (sbi->s_flex_groups_allocated *
1973 sizeof(struct flex_groups)));
b93b41d4 1974 kvfree(sbi->s_flex_groups);
117fff10
TT		 1975 }
1976 sbi->s_flex_groups = new_groups;
1977 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1978 return 0;
1979}
1980
772cb7c8
JS		 1981static int ext4_fill_flex_info(struct super_block *sb)
1982{
1983 struct ext4_sb_info *sbi = EXT4_SB(sb);
1984 struct ext4_group_desc *gdp = NULL;
772cb7c8 1985 ext4_group_t flex_group;
117fff10 1986 int i, err;
772cb7c8 1987
503358ae 1988 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1989 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1990 sbi->s_log_groups_per_flex = 0;
1991 return 1;
1992 }
1993
117fff10
TT		 1994 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1995 if (err)
9933fc0a 1996 goto failed;
772cb7c8 1997
772cb7c8 1998 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1999 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2000
2001 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2002 atomic_add(ext4_free_inodes_count(sb, gdp),
2003 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2004 atomic64_add(ext4_free_group_clusters(sb, gdp),
2005 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2006 atomic_add(ext4_used_dirs_count(sb, gdp),
2007 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2008 }
2009
2010 return 1;
2011failed:
2012 return 0;
2013}
2014
feb0ab32
DW		 2015static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2016 struct ext4_group_desc *gdp)
717d50e4 2017{
feb0ab32 2018 int offset;
717d50e4 2019 __u16 crc = 0;
feb0ab32 2020 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2021
9aa5d32b 2022 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2023 /* Use new metadata_csum algorithm */
d6a77105 2024 __le16 save_csum;
feb0ab32
DW		 2025 __u32 csum32;
2026
d6a77105 2027 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2028 gdp->bg_checksum = 0;
2029 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2030 sizeof(le_group));
2031 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2032 sbi->s_desc_size);
d6a77105 2033 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2034
2035 crc = csum32 & 0xFFFF;
2036 goto out;
717d50e4
AD		 2037 }
2038
feb0ab32 2039 /* old crc16 code */
813d32f9
DW		 2040 if (!(sbi->s_es->s_feature_ro_compat &
2041 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2042 return 0;
2043
feb0ab32
DW		 2044 offset = offsetof(struct ext4_group_desc, bg_checksum);
2045
2046 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2047 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2048 crc = crc16(crc, (__u8 *)gdp, offset);
2049 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2050 /* for checksum of struct ext4_group_desc do the rest...*/
2051 if ((sbi->s_es->s_feature_incompat &
2052 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2053 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2054 crc = crc16(crc, (__u8 *)gdp + offset,
2055 le16_to_cpu(sbi->s_es->s_desc_size) -
2056 offset);
2057
2058out:
717d50e4
AD		 2059 return cpu_to_le16(crc);
2060}
2061
feb0ab32 2062int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2063 struct ext4_group_desc *gdp)
2064{
feb0ab32
DW		 2065 if (ext4_has_group_desc_csum(sb) &&
2066 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2067 block_group, gdp)))
717d50e4
AD		 2068 return 0;
2069
2070 return 1;
2071}
2072
feb0ab32
DW		 2073void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2074 struct ext4_group_desc *gdp)
2075{
2076 if (!ext4_has_group_desc_csum(sb))
2077 return;
2078 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2079}
2080
ac27a0ec 2081/* Called at mount-time, super-block is locked */
bfff6873
LC		 2082static int ext4_check_descriptors(struct super_block *sb,
2083 ext4_group_t *first_not_zeroed)
ac27a0ec 2084{
617ba13b
MC		 2085 struct ext4_sb_info *sbi = EXT4_SB(sb);
2086 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2087 ext4_fsblk_t last_block;
bd81d8ee
LV		 2088 ext4_fsblk_t block_bitmap;
2089 ext4_fsblk_t inode_bitmap;
2090 ext4_fsblk_t inode_table;
ce421581 2091 int flexbg_flag = 0;
bfff6873 2092 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2093
ce421581
JS		 2094 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2095 flexbg_flag = 1;
2096
af5bc92d 2097 ext4_debug("Checking group descriptors");
ac27a0ec 2098
197cd65a
AM		 2099 for (i = 0; i < sbi->s_groups_count; i++) {
2100 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2101
ce421581 2102 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2103 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2104 else
2105 last_block = first_block +
617ba13b 2106 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2107
bfff6873
LC		 2108 if ((grp == sbi->s_groups_count) &&
2109 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2110 grp = i;
2111
8fadc143 2112 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2113 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2114 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2115 "Block bitmap for group %u not in group "
b31e1552 2116 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2117 return 0;
2118 }
8fadc143 2119 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2120 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2121 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2122 "Inode bitmap for group %u not in group "
b31e1552 2123 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2124 return 0;
2125 }
8fadc143 2126 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2127 if (inode_table < first_block ||
2b2d6d01 2128 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2129 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2130 "Inode table for group %u not in group "
b31e1552 2131 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2132 return 0;
2133 }
955ce5f5 2134 ext4_lock_group(sb, i);
feb0ab32 2135 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2136 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2137 "Checksum for group %u failed (%u!=%u)",
2138 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2139 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2140 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2141 ext4_unlock_group(sb, i);
8a266467 2142 return 0;
7ee1ec4c 2143 }
717d50e4 2144 }
955ce5f5 2145 ext4_unlock_group(sb, i);
ce421581
JS		 2146 if (!flexbg_flag)
2147 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2148 }
bfff6873
LC		 2149 if (NULL != first_not_zeroed)
2150 *first_not_zeroed = grp;
ac27a0ec
DK		 2151 return 1;
2152}
2153
617ba13b 2154/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2155 * the superblock) which were deleted from all directories, but held open by
2156 * a process at the time of a crash. We walk the list and try to delete these
2157 * inodes at recovery time (only with a read-write filesystem).
2158 *
2159 * In order to keep the orphan inode chain consistent during traversal (in
2160 * case of crash during recovery), we link each inode into the superblock
2161 * orphan list_head and handle it the same way as an inode deletion during
2162 * normal operation (which journals the operations for us).
2163 *
2164 * We only do an iget() and an iput() on each inode, which is very safe if we
2165 * accidentally point at an in-use or already deleted inode. The worst that
2166 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2167 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2168 * e2fsck was run on this filesystem, and it must have already done the orphan
2169 * inode cleanup for us, so we can safely abort without any further action.
2170 */
2b2d6d01
TT		 2171static void ext4_orphan_cleanup(struct super_block *sb,
2172 struct ext4_super_block *es)
ac27a0ec
DK		 2173{
2174 unsigned int s_flags = sb->s_flags;
2175 int nr_orphans = 0, nr_truncates = 0;
2176#ifdef CONFIG_QUOTA
2177 int i;
2178#endif
2179 if (!es->s_last_orphan) {
2180 jbd_debug(4, "no orphan inodes to clean up\n");
2181 return;
2182 }
2183
a8f48a95 2184 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2185 ext4_msg(sb, KERN_ERR, "write access "
2186 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2187 return;
2188 }
2189
d39195c3
AG		 2190 /* Check if feature set would not allow a r/w mount */
2191 if (!ext4_feature_set_ok(sb, 0)) {
2192 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2193 "unknown ROCOMPAT features");
2194 return;
2195 }
2196
617ba13b 2197 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2198 /* don't clear list on RO mount w/ errors */
2199 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
84474976 2200 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 2201 "clearing orphan list.\n");
c25f9bc6
ES		 2202 es->s_last_orphan = 0;
2203 }
ac27a0ec
DK		 2204 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2205 return;
2206 }
2207
2208 if (s_flags & MS_RDONLY) {
b31e1552 2209 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2210 sb->s_flags &= ~MS_RDONLY;
2211 }
2212#ifdef CONFIG_QUOTA
2213 /* Needed for iput() to work correctly and not trash data */
2214 sb->s_flags |= MS_ACTIVE;
2215 /* Turn on quotas so that they are updated correctly */
a2d4a646 2216 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 2217 if (EXT4_SB(sb)->s_qf_names[i]) {
2218 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2219 if (ret < 0)
b31e1552
ES		 2220 ext4_msg(sb, KERN_ERR,
2221 "Cannot turn on journaled "
2222 "quota: error %d", ret);
ac27a0ec
DK		 2223 }
2224 }
2225#endif
2226
2227 while (es->s_last_orphan) {
2228 struct inode *inode;
2229
97bd42b9
JB		 2230 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2231 if (IS_ERR(inode)) {
ac27a0ec
DK		 2232 es->s_last_orphan = 0;
2233 break;
2234 }
2235
617ba13b 2236 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2237 dquot_initialize(inode);
ac27a0ec 2238 if (inode->i_nlink) {
566370a2
PT		 2239 if (test_opt(sb, DEBUG))
2240 ext4_msg(sb, KERN_DEBUG,
2241 "%s: truncating inode %lu to %lld bytes",
2242 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2243 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2244 inode->i_ino, inode->i_size);
721e3eba 2245 mutex_lock(&inode->i_mutex);
55f252c9 2246 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2247 ext4_truncate(inode);
721e3eba 2248 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2249 nr_truncates++;
2250 } else {
566370a2
PT		 2251 if (test_opt(sb, DEBUG))
2252 ext4_msg(sb, KERN_DEBUG,
2253 "%s: deleting unreferenced inode %lu",
2254 __func__, inode->i_ino);
ac27a0ec
DK		 2255 jbd_debug(2, "deleting unreferenced inode %lu\n",
2256 inode->i_ino);
2257 nr_orphans++;
2258 }
2259 iput(inode); /* The delete magic happens here! */
2260 }
2261
2b2d6d01 2262#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2263
2264 if (nr_orphans)
b31e1552
ES		 2265 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2266 PLURAL(nr_orphans));
ac27a0ec 2267 if (nr_truncates)
b31e1552
ES		 2268 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2269 PLURAL(nr_truncates));
ac27a0ec
DK		 2270#ifdef CONFIG_QUOTA
2271 /* Turn quotas off */
a2d4a646 2272 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
ac27a0ec 2273 if (sb_dqopt(sb)->files[i])
287a8095 2274 dquot_quota_off(sb, i);
ac27a0ec
DK		 2275 }
2276#endif
2277 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2278}
0b8e58a1 2279
cd2291a4
ES		 2280/*
2281 * Maximal extent format file size.
2282 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2283 * extent format containers, within a sector_t, and within i_blocks
2284 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2285 * so that won't be a limiting factor.
2286 *
f17722f9
LC		 2287 * However there is other limiting factor. We do store extents in the form
2288 * of starting block and length, hence the resulting length of the extent
2289 * covering maximum file size must fit into on-disk format containers as
2290 * well. Given that length is always by 1 unit bigger than max unit (because
2291 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2292 *
cd2291a4
ES		 2293 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2294 */
f287a1a5 2295static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2296{
2297 loff_t res;
2298 loff_t upper_limit = MAX_LFS_FILESIZE;
2299
2300 /* small i_blocks in vfs inode? */
f287a1a5 2301 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2302 /*
90c699a9 2303 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2304 * i_block represent total blocks in 512 bytes
2305 * 32 == size of vfs inode i_blocks * 8
2306 */
2307 upper_limit = (1LL << 32) - 1;
2308
2309 /* total blocks in file system block size */
2310 upper_limit >>= (blkbits - 9);
2311 upper_limit <<= blkbits;
2312 }
2313
f17722f9
LC		 2314 /*
2315 * 32-bit extent-start container, ee_block. We lower the maxbytes
2316 * by one fs block, so ee_len can cover the extent of maximum file
2317 * size
2318 */
2319 res = (1LL << 32) - 1;
cd2291a4 2320 res <<= blkbits;
cd2291a4
ES		 2321
2322 /* Sanity check against vm- & vfs- imposed limits */
2323 if (res > upper_limit)
2324 res = upper_limit;
2325
2326 return res;
2327}
ac27a0ec 2328
ac27a0ec 2329/*
cd2291a4 2330 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2331 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2332 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2333 */
f287a1a5 2334static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2335{
617ba13b 2336 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2337 int meta_blocks;
2338 loff_t upper_limit;
0b8e58a1
AD		 2339 /* This is calculated to be the largest file size for a dense, block
2340 * mapped file such that the file's total number of 512-byte sectors,
2341 * including data and all indirect blocks, does not exceed (2^48 - 1).
2342 *
2343 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2344 * number of 512-byte sectors of the file.
0fc1b451
AK		 2345 */
2346
f287a1a5 2347 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2348 /*
90c699a9 2349 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2350 * the inode i_block field represents total file blocks in
2351 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2352 */
2353 upper_limit = (1LL << 32) - 1;
2354
2355 /* total blocks in file system block size */
2356 upper_limit >>= (bits - 9);
2357
2358 } else {
8180a562
AK		 2359 /*
2360 * We use 48 bit ext4_inode i_blocks
2361 * With EXT4_HUGE_FILE_FL set the i_blocks
2362 * represent total number of blocks in
2363 * file system block size
2364 */
0fc1b451
AK		 2365 upper_limit = (1LL << 48) - 1;
2366
0fc1b451
AK		 2367 }
2368
2369 /* indirect blocks */
2370 meta_blocks = 1;
2371 /* double indirect blocks */
2372 meta_blocks += 1 + (1LL << (bits-2));
2373 /* tripple indirect blocks */
2374 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2375
2376 upper_limit -= meta_blocks;
2377 upper_limit <<= bits;
ac27a0ec
DK		 2378
2379 res += 1LL << (bits-2);
2380 res += 1LL << (2*(bits-2));
2381 res += 1LL << (3*(bits-2));
2382 res <<= bits;
2383 if (res > upper_limit)
2384 res = upper_limit;
0fc1b451
AK		 2385
2386 if (res > MAX_LFS_FILESIZE)
2387 res = MAX_LFS_FILESIZE;
2388
ac27a0ec
DK		 2389 return res;
2390}
2391
617ba13b 2392static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2393 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2394{
617ba13b 2395 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2396 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2397 int has_super = 0;
2398
2399 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2400
617ba13b 2401 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2402 nr < first_meta_bg)
70bbb3e0 2403 return logical_sb_block + nr + 1;
ac27a0ec 2404 bg = sbi->s_desc_per_block * nr;
617ba13b 2405 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2406 has_super = 1;
0b8e58a1 2407
bd63f6b0
DW		 2408 /*
2409 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2410 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2411 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2412 * compensate.
2413 */
2414 if (sb->s_blocksize == 1024 && nr == 0 &&
2415 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2416 has_super++;
2417
617ba13b 2418 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2419}
2420
c9de560d
AT		 2421/**
2422 * ext4_get_stripe_size: Get the stripe size.
2423 * @sbi: In memory super block info
2424 *
2425 * If we have specified it via mount option, then
2426 * use the mount option value. If the value specified at mount time is
2427 * greater than the blocks per group use the super block value.
2428 * If the super block value is greater than blocks per group return 0.
2429 * Allocator needs it be less than blocks per group.
2430 *
2431 */
2432static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2433{
2434 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2435 unsigned long stripe_width =
2436 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2437 int ret;
c9de560d
AT		 2438
2439 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2440 ret = sbi->s_stripe;
2441 else if (stripe_width <= sbi->s_blocks_per_group)
2442 ret = stripe_width;
2443 else if (stride <= sbi->s_blocks_per_group)
2444 ret = stride;
2445 else
2446 ret = 0;
c9de560d 2447
3eb08658
DE		 2448 /*
2449 * If the stripe width is 1, this makes no sense and
2450 * we set it to 0 to turn off stripe handling code.
2451 */
2452 if (ret <= 1)
2453 ret = 0;
c9de560d 2454
3eb08658 2455 return ret;
c9de560d 2456}
ac27a0ec 2457
3197ebdb
TT		 2458/* sysfs supprt */
2459
2460struct ext4_attr {
2461 struct attribute attr;
2462 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2463 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2464 const char *, size_t);
f2d50a65
JK		 2465 union {
2466 int offset;
2467 int deprecated_val;
2468 } u;
3197ebdb
TT		 2469};
2470
27dd4385
LC		 2471static int parse_strtoull(const char *buf,
2472 unsigned long long max, unsigned long long *value)
2473{
2474 int ret;
2475
2476 ret = kstrtoull(skip_spaces(buf), 0, value);
2477 if (!ret && *value > max)
2478 ret = -EINVAL;
2479 return ret;
2480}
2481
3197ebdb
TT		 2482static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi,
2484 char *buf)
2485{
2486 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2487 (s64) EXT4_C2B(sbi,
2488 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2489}
2490
2491static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2492 struct ext4_sb_info *sbi, char *buf)
2493{
2494 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2495
f613dfcb
TT		 2496 if (!sb->s_bdev->bd_part)
2497 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2498 return snprintf(buf, PAGE_SIZE, "%lu\n",
2499 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2500 sbi->s_sectors_written_start) >> 1);
2501}
2502
2503static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2504 struct ext4_sb_info *sbi, char *buf)
2505{
2506 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2507
f613dfcb
TT		 2508 if (!sb->s_bdev->bd_part)
2509 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2510 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2511 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2512 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2513 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2514}
2515
2516static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2519{
2520 unsigned long t;
e1091b15 2521 int ret;
3197ebdb 2522
e1091b15
LC		 2523 ret = kstrtoul(skip_spaces(buf), 0, &t);
2524 if (ret)
2525 return ret;
3197ebdb 2526
e1091b15 2527 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2528 return -EINVAL;
2529
2530 sbi->s_inode_readahead_blks = t;
2531 return count;
2532}
2533
2534static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2535 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2536{
f2d50a65 2537 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2538
2539 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2540}
2541
2542static ssize_t sbi_ui_store(struct ext4_attr *a,
2543 struct ext4_sb_info *sbi,
2544 const char *buf, size_t count)
2545{
f2d50a65 2546 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2547 unsigned long t;
e1091b15 2548 int ret;
3197ebdb 2549
e1091b15
LC		 2550 ret = kstrtoul(skip_spaces(buf), 0, &t);
2551 if (ret)
2552 return ret;
3197ebdb
TT		 2553 *ui = t;
2554 return count;
2555}
2556
52c198c6
LC		 2557static ssize_t es_ui_show(struct ext4_attr *a,
2558 struct ext4_sb_info *sbi, char *buf)
2559{
2560
2561 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2562 a->u.offset);
2563
2564 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2565}
2566
27dd4385
LC		 2567static ssize_t reserved_clusters_show(struct ext4_attr *a,
2568 struct ext4_sb_info *sbi, char *buf)
2569{
2570 return snprintf(buf, PAGE_SIZE, "%llu\n",
2571 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2572}
2573
2574static ssize_t reserved_clusters_store(struct ext4_attr *a,
2575 struct ext4_sb_info *sbi,
2576 const char *buf, size_t count)
2577{
2578 unsigned long long val;
2579 int ret;
2580
2581 if (parse_strtoull(buf, -1ULL, &val))
2582 return -EINVAL;
2583 ret = ext4_reserve_clusters(sbi, val);
2584
2585 return ret ? ret : count;
2586}
2587
2c0544b2
TT		 2588static ssize_t trigger_test_error(struct ext4_attr *a,
2589 struct ext4_sb_info *sbi,
2590 const char *buf, size_t count)
2591{
2592 int len = count;
2593
2594 if (!capable(CAP_SYS_ADMIN))
2595 return -EPERM;
2596
2597 if (len && buf[len-1] == '\n')
2598 len--;
2599
2600 if (len)
2601 ext4_error(sbi->s_sb, "%.*s", len, buf);
2602 return count;
2603}
2604
f2d50a65
JK		 2605static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2606 struct ext4_sb_info *sbi, char *buf)
2607{
2608 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2609}
2610
3197ebdb
TT		 2611#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2612static struct ext4_attr ext4_attr_##_name = { \
2613 .attr = {.name = __stringify(_name), .mode = _mode }, \
2614 .show = _show, \
2615 .store = _store, \
f2d50a65
JK		 2616 .u = { \
2617 .offset = offsetof(struct ext4_sb_info, _elname),\
2618 }, \
3197ebdb 2619}
52c198c6
LC		 2620
2621#define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2622static struct ext4_attr ext4_attr_##_name = { \
2623 .attr = {.name = __stringify(_name), .mode = _mode }, \
2624 .show = _show, \
2625 .store = _store, \
2626 .u = { \
2627 .offset = offsetof(struct ext4_super_block, _elname), \
2628 }, \
2629}
2630
3197ebdb
TT		 2631#define EXT4_ATTR(name, mode, show, store) \
2632static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2633
857ac889 2634#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2635#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2636#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
52c198c6
LC		 2637
2638#define EXT4_RO_ATTR_ES_UI(name, elname) \
2639 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
3197ebdb
TT		 2640#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2641 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
52c198c6 2642
3197ebdb 2643#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2644#define EXT4_DEPRECATED_ATTR(_name, _val) \
2645static struct ext4_attr ext4_attr_##_name = { \
2646 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2647 .show = sbi_deprecated_show, \
2648 .u = { \
2649 .deprecated_val = _val, \
2650 }, \
2651}
3197ebdb
TT		 2652
2653EXT4_RO_ATTR(delayed_allocation_blocks);
2654EXT4_RO_ATTR(session_write_kbytes);
2655EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2656EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2657EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2658 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2659EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2660EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2661EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2662EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2663EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2664EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2665EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2666EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2667EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2668EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2669EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2670EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2671EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2672EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2673EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2674EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
52c198c6
LC		 2675EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2676EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2677EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
3197ebdb
TT		 2678
2679static struct attribute *ext4_attrs[] = {
2680 ATTR_LIST(delayed_allocation_blocks),
2681 ATTR_LIST(session_write_kbytes),
2682 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2683 ATTR_LIST(reserved_clusters),
3197ebdb 2684 ATTR_LIST(inode_readahead_blks),
11013911 2685 ATTR_LIST(inode_goal),
3197ebdb
TT		 2686 ATTR_LIST(mb_stats),
2687 ATTR_LIST(mb_max_to_scan),
2688 ATTR_LIST(mb_min_to_scan),
2689 ATTR_LIST(mb_order2_req),
2690 ATTR_LIST(mb_stream_req),
2691 ATTR_LIST(mb_group_prealloc),
55138e0b 2692 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2693 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2694 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2695 ATTR_LIST(err_ratelimit_interval_ms),
2696 ATTR_LIST(err_ratelimit_burst),
2697 ATTR_LIST(warning_ratelimit_interval_ms),
2698 ATTR_LIST(warning_ratelimit_burst),
2699 ATTR_LIST(msg_ratelimit_interval_ms),
2700 ATTR_LIST(msg_ratelimit_burst),
52c198c6
LC		 2701 ATTR_LIST(errors_count),
2702 ATTR_LIST(first_error_time),
2703 ATTR_LIST(last_error_time),
3197ebdb
TT		 2704 NULL,
2705};
2706
857ac889
LC		 2707/* Features this copy of ext4 supports */
2708EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2709EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2710EXT4_INFO_ATTR(meta_bg_resize);
6ddb2447 2711EXT4_INFO_ATTR(encryption);
857ac889
LC		 2712
2713static struct attribute *ext4_feat_attrs[] = {
2714 ATTR_LIST(lazy_itable_init),
27ee40df 2715 ATTR_LIST(batched_discard),
5e7bbef1 2716 ATTR_LIST(meta_bg_resize),
6ddb2447 2717 ATTR_LIST(encryption),
857ac889
LC		 2718 NULL,
2719};
2720
3197ebdb
TT		 2721static ssize_t ext4_attr_show(struct kobject *kobj,
2722 struct attribute *attr, char *buf)
2723{
2724 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2725 s_kobj);
2726 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2727
2728 return a->show ? a->show(a, sbi, buf) : 0;
2729}
2730
2731static ssize_t ext4_attr_store(struct kobject *kobj,
2732 struct attribute *attr,
2733 const char *buf, size_t len)
2734{
2735 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2736 s_kobj);
2737 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2738
2739 return a->store ? a->store(a, sbi, buf, len) : 0;
2740}
2741
2742static void ext4_sb_release(struct kobject *kobj)
2743{
2744 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2745 s_kobj);
2746 complete(&sbi->s_kobj_unregister);
2747}
2748
52cf25d0 2749static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2750 .show = ext4_attr_show,
2751 .store = ext4_attr_store,
2752};
2753
2754static struct kobj_type ext4_ktype = {
2755 .default_attrs = ext4_attrs,
2756 .sysfs_ops = &ext4_attr_ops,
2757 .release = ext4_sb_release,
2758};
2759
857ac889
LC		 2760static void ext4_feat_release(struct kobject *kobj)
2761{
2762 complete(&ext4_feat->f_kobj_unregister);
2763}
2764
c7f72543
LC		 2765static ssize_t ext4_feat_show(struct kobject *kobj,
2766 struct attribute *attr, char *buf)
2767{
2768 return snprintf(buf, PAGE_SIZE, "supported\n");
2769}
2770
2771/*
2772 * We can not use ext4_attr_show/store because it relies on the kobject
2773 * being embedded in the ext4_sb_info structure which is definitely not
2774 * true in this case.
2775 */
2776static const struct sysfs_ops ext4_feat_ops = {
2777 .show = ext4_feat_show,
2778 .store = NULL,
2779};
2780
857ac889
LC		 2781static struct kobj_type ext4_feat_ktype = {
2782 .default_attrs = ext4_feat_attrs,
c7f72543 2783 .sysfs_ops = &ext4_feat_ops,
857ac889
LC		 2784 .release = ext4_feat_release,
2785};
2786
a13fb1a4
ES		 2787/*
2788 * Check whether this filesystem can be mounted based on
2789 * the features present and the RDONLY/RDWR mount requested.
2790 * Returns 1 if this filesystem can be mounted as requested,
2791 * 0 if it cannot be.
2792 */
2793static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2794{
2795 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2796 ext4_msg(sb, KERN_ERR,
2797 "Couldn't mount because of "
2798 "unsupported optional features (%x)",
2799 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2800 ~EXT4_FEATURE_INCOMPAT_SUPP));
2801 return 0;
2802 }
2803
2804 if (readonly)
2805 return 1;
2806
2cb5cc8b
DW		 2807 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2808 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2809 sb->s_flags |= MS_RDONLY;
2810 return 1;
2811 }
2812
a13fb1a4
ES		 2813 /* Check that feature set is OK for a read-write mount */
2814 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2815 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2816 "unsupported optional features (%x)",
2817 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2818 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2819 return 0;
2820 }
2821 /*
2822 * Large file size enabled file system can only be mounted
2823 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2824 */
2825 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2826 if (sizeof(blkcnt_t) < sizeof(u64)) {
2827 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2828 "cannot be mounted RDWR without "
2829 "CONFIG_LBDAF");
2830 return 0;
2831 }
2832 }
bab08ab9
TT		 2833 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2834 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2835 ext4_msg(sb, KERN_ERR,
2836 "Can't support bigalloc feature without "
2837 "extents feature\n");
2838 return 0;
2839 }
7c319d32
AK		 2840
2841#ifndef CONFIG_QUOTA
2842 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2843 !readonly) {
2844 ext4_msg(sb, KERN_ERR,
2845 "Filesystem with quota feature cannot be mounted RDWR "
2846 "without CONFIG_QUOTA");
2847 return 0;
2848 }
2849#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2850 return 1;
2851}
2852
66e61a9e
TT		 2853/*
2854 * This function is called once a day if we have errors logged
2855 * on the file system
2856 */
2857static void print_daily_error_info(unsigned long arg)
2858{
2859 struct super_block *sb = (struct super_block *) arg;
2860 struct ext4_sb_info *sbi;
2861 struct ext4_super_block *es;
2862
2863 sbi = EXT4_SB(sb);
2864 es = sbi->s_es;
2865
2866 if (es->s_error_count)
ae0f78de
TT		 2867 /* fsck newer than v1.41.13 is needed to clean this condition. */
2868 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2869 le32_to_cpu(es->s_error_count));
2870 if (es->s_first_error_time) {
ae0f78de 2871 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2872 sb->s_id, le32_to_cpu(es->s_first_error_time),
2873 (int) sizeof(es->s_first_error_func),
2874 es->s_first_error_func,
2875 le32_to_cpu(es->s_first_error_line));
2876 if (es->s_first_error_ino)
2877 printk(": inode %u",
2878 le32_to_cpu(es->s_first_error_ino));
2879 if (es->s_first_error_block)
2880 printk(": block %llu", (unsigned long long)
2881 le64_to_cpu(es->s_first_error_block));
2882 printk("\n");
2883 }
2884 if (es->s_last_error_time) {
ae0f78de 2885 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2886 sb->s_id, le32_to_cpu(es->s_last_error_time),
2887 (int) sizeof(es->s_last_error_func),
2888 es->s_last_error_func,
2889 le32_to_cpu(es->s_last_error_line));
2890 if (es->s_last_error_ino)
2891 printk(": inode %u",
2892 le32_to_cpu(es->s_last_error_ino));
2893 if (es->s_last_error_block)
2894 printk(": block %llu", (unsigned long long)
2895 le64_to_cpu(es->s_last_error_block));
2896 printk("\n");
2897 }
2898 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2899}
2900
bfff6873
LC		 2901/* Find next suitable group and run ext4_init_inode_table */
2902static int ext4_run_li_request(struct ext4_li_request *elr)
2903{
2904 struct ext4_group_desc *gdp = NULL;
2905 ext4_group_t group, ngroups;
2906 struct super_block *sb;
2907 unsigned long timeout = 0;
2908 int ret = 0;
2909
2910 sb = elr->lr_super;
2911 ngroups = EXT4_SB(sb)->s_groups_count;
2912
8e8ad8a5 2913 sb_start_write(sb);
bfff6873
LC		 2914 for (group = elr->lr_next_group; group < ngroups; group++) {
2915 gdp = ext4_get_group_desc(sb, group, NULL);
2916 if (!gdp) {
2917 ret = 1;
2918 break;
2919 }
2920
2921 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2922 break;
2923 }
2924
7f511862 2925 if (group >= ngroups)
bfff6873
LC		 2926 ret = 1;
2927
2928 if (!ret) {
2929 timeout = jiffies;
2930 ret = ext4_init_inode_table(sb, group,
2931 elr->lr_timeout ? 0 : 1);
2932 if (elr->lr_timeout == 0) {
51ce6511
LC		 2933 timeout = (jiffies - timeout) *
2934 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2935 elr->lr_timeout = timeout;
2936 }
2937 elr->lr_next_sched = jiffies + elr->lr_timeout;
2938 elr->lr_next_group = group + 1;
2939 }
8e8ad8a5 2940 sb_end_write(sb);
bfff6873
LC		 2941
2942 return ret;
2943}
2944
2945/*
2946 * Remove lr_request from the list_request and free the
4ed5c033 2947 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2948 */
2949static void ext4_remove_li_request(struct ext4_li_request *elr)
2950{
2951 struct ext4_sb_info *sbi;
2952
2953 if (!elr)
2954 return;
2955
2956 sbi = elr->lr_sbi;
2957
2958 list_del(&elr->lr_request);
2959 sbi->s_li_request = NULL;
2960 kfree(elr);
2961}
2962
2963static void ext4_unregister_li_request(struct super_block *sb)
2964{
1bb933fb
LC		 2965 mutex_lock(&ext4_li_mtx);
2966 if (!ext4_li_info) {
2967 mutex_unlock(&ext4_li_mtx);
bfff6873 2968 return;
1bb933fb 2969 }
bfff6873
LC		 2970
2971 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2972 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2973 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2974 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2975}
2976
8f1f7453
ES		 2977static struct task_struct *ext4_lazyinit_task;
2978
bfff6873
LC		 2979/*
2980 * This is the function where ext4lazyinit thread lives. It walks
2981 * through the request list searching for next scheduled filesystem.
2982 * When such a fs is found, run the lazy initialization request
2983 * (ext4_rn_li_request) and keep track of the time spend in this
2984 * function. Based on that time we compute next schedule time of
2985 * the request. When walking through the list is complete, compute
2986 * next waking time and put itself into sleep.
2987 */
2988static int ext4_lazyinit_thread(void *arg)
2989{
2990 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2991 struct list_head *pos, *n;
2992 struct ext4_li_request *elr;
4ed5c033 2993 unsigned long next_wakeup, cur;
bfff6873
LC		 2994
2995 BUG_ON(NULL == eli);
2996
bfff6873
LC		 2997cont_thread:
2998 while (true) {
2999 next_wakeup = MAX_JIFFY_OFFSET;
3000
3001 mutex_lock(&eli->li_list_mtx);
3002 if (list_empty(&eli->li_request_list)) {
3003 mutex_unlock(&eli->li_list_mtx);
3004 goto exit_thread;
3005 }
3006
3007 list_for_each_safe(pos, n, &eli->li_request_list) {
3008 elr = list_entry(pos, struct ext4_li_request,
3009 lr_request);
3010
b2c78cd0
TT		 3011 if (time_after_eq(jiffies, elr->lr_next_sched)) {
3012 if (ext4_run_li_request(elr) != 0) {
3013 /* error, remove the lazy_init job */
3014 ext4_remove_li_request(elr);
3015 continue;
3016 }
bfff6873
LC		 3017 }
3018
3019 if (time_before(elr->lr_next_sched, next_wakeup))
3020 next_wakeup = elr->lr_next_sched;
3021 }
3022 mutex_unlock(&eli->li_list_mtx);
3023
a0acae0e 3024 try_to_freeze();
bfff6873 3025
4ed5c033
LC		 3026 cur = jiffies;
3027 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3028 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3029 cond_resched();
3030 continue;
3031 }
3032
4ed5c033
LC		 3033 schedule_timeout_interruptible(next_wakeup - cur);
3034
8f1f7453
ES		 3035 if (kthread_should_stop()) {
3036 ext4_clear_request_list();
3037 goto exit_thread;
3038 }
bfff6873
LC		 3039 }
3040
3041exit_thread:
3042 /*
3043 * It looks like the request list is empty, but we need
3044 * to check it under the li_list_mtx lock, to prevent any
3045 * additions into it, and of course we should lock ext4_li_mtx
3046 * to atomically free the list and ext4_li_info, because at
3047 * this point another ext4 filesystem could be registering
3048 * new one.
3049 */
3050 mutex_lock(&ext4_li_mtx);
3051 mutex_lock(&eli->li_list_mtx);
3052 if (!list_empty(&eli->li_request_list)) {
3053 mutex_unlock(&eli->li_list_mtx);
3054 mutex_unlock(&ext4_li_mtx);
3055 goto cont_thread;
3056 }
3057 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3058 kfree(ext4_li_info);
3059 ext4_li_info = NULL;
3060 mutex_unlock(&ext4_li_mtx);
3061
3062 return 0;
3063}
3064
3065static void ext4_clear_request_list(void)
3066{
3067 struct list_head *pos, *n;
3068 struct ext4_li_request *elr;
3069
3070 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3071 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3072 elr = list_entry(pos, struct ext4_li_request,
3073 lr_request);
3074 ext4_remove_li_request(elr);
3075 }
3076 mutex_unlock(&ext4_li_info->li_list_mtx);
3077}
3078
3079static int ext4_run_lazyinit_thread(void)
3080{
8f1f7453
ES		 3081 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3082 ext4_li_info, "ext4lazyinit");
3083 if (IS_ERR(ext4_lazyinit_task)) {
3084 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3085 ext4_clear_request_list();
bfff6873
LC		 3086 kfree(ext4_li_info);
3087 ext4_li_info = NULL;
92b97816 3088 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3089 "initialization thread\n",
3090 err);
3091 return err;
3092 }
3093 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3094 return 0;
3095}
3096
3097/*
3098 * Check whether it make sense to run itable init. thread or not.
3099 * If there is at least one uninitialized inode table, return
3100 * corresponding group number, else the loop goes through all
3101 * groups and return total number of groups.
3102 */
3103static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3104{
3105 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3106 struct ext4_group_desc *gdp = NULL;
3107
3108 for (group = 0; group < ngroups; group++) {
3109 gdp = ext4_get_group_desc(sb, group, NULL);
3110 if (!gdp)
3111 continue;
3112
3113 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3114 break;
3115 }
3116
3117 return group;
3118}
3119
3120static int ext4_li_info_new(void)
3121{
3122 struct ext4_lazy_init *eli = NULL;
3123
3124 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3125 if (!eli)
3126 return -ENOMEM;
3127
bfff6873
LC		 3128 INIT_LIST_HEAD(&eli->li_request_list);
3129 mutex_init(&eli->li_list_mtx);
3130
bfff6873
LC		 3131 eli->li_state |= EXT4_LAZYINIT_QUIT;
3132
3133 ext4_li_info = eli;
3134
3135 return 0;
3136}
3137
3138static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3139 ext4_group_t start)
3140{
3141 struct ext4_sb_info *sbi = EXT4_SB(sb);
3142 struct ext4_li_request *elr;
bfff6873
LC		 3143
3144 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3145 if (!elr)
3146 return NULL;
3147
3148 elr->lr_super = sb;
3149 elr->lr_sbi = sbi;
3150 elr->lr_next_group = start;
3151
3152 /*
3153 * Randomize first schedule time of the request to
3154 * spread the inode table initialization requests
3155 * better.
3156 */
dd1f723b
TT		 3157 elr->lr_next_sched = jiffies + (prandom_u32() %
3158 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3159 return elr;
3160}
3161
7f511862
TT		 3162int ext4_register_li_request(struct super_block *sb,
3163 ext4_group_t first_not_zeroed)
bfff6873
LC		 3164{
3165 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3166 struct ext4_li_request *elr = NULL;
bfff6873 3167 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3168 int ret = 0;
bfff6873 3169
7f511862 3170 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3171 if (sbi->s_li_request != NULL) {
3172 /*
3173 * Reset timeout so it can be computed again, because
3174 * s_li_wait_mult might have changed.
3175 */
3176 sbi->s_li_request->lr_timeout = 0;
7f511862 3177 goto out;
51ce6511 3178 }
bfff6873
LC		 3179
3180 if (first_not_zeroed == ngroups ||
3181 (sb->s_flags & MS_RDONLY) ||
55ff3840 3182 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3183 goto out;
bfff6873
LC		 3184
3185 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3186 if (!elr) {
3187 ret = -ENOMEM;
3188 goto out;
3189 }
bfff6873
LC		 3190
3191 if (NULL == ext4_li_info) {
3192 ret = ext4_li_info_new();
3193 if (ret)
3194 goto out;
3195 }
3196
3197 mutex_lock(&ext4_li_info->li_list_mtx);
3198 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3199 mutex_unlock(&ext4_li_info->li_list_mtx);
3200
3201 sbi->s_li_request = elr;
46e4690b
TM		 3202 /*
3203 * set elr to NULL here since it has been inserted to
3204 * the request_list and the removal and free of it is
3205 * handled by ext4_clear_request_list from now on.
3206 */
3207 elr = NULL;
bfff6873
LC		 3208
3209 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3210 ret = ext4_run_lazyinit_thread();
3211 if (ret)
3212 goto out;
3213 }
bfff6873 3214out:
beed5ecb
NK		 3215 mutex_unlock(&ext4_li_mtx);
3216 if (ret)
bfff6873 3217 kfree(elr);
bfff6873
LC		 3218 return ret;
3219}
3220
3221/*
3222 * We do not need to lock anything since this is called on
3223 * module unload.
3224 */
3225static void ext4_destroy_lazyinit_thread(void)
3226{
3227 /*
3228 * If thread exited earlier
3229 * there's nothing to be done.
3230 */
8f1f7453 3231 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3232 return;
3233
8f1f7453 3234 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3235}
3236
25ed6e8a
DW		 3237static int set_journal_csum_feature_set(struct super_block *sb)
3238{
3239 int ret = 1;
3240 int compat, incompat;
3241 struct ext4_sb_info *sbi = EXT4_SB(sb);
3242
9aa5d32b 3243 if (ext4_has_metadata_csum(sb)) {
db9ee220 3244 /* journal checksum v3 */
25ed6e8a 3245 compat = 0;
db9ee220 3246 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3247 } else {
3248 /* journal checksum v1 */
3249 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3250 incompat = 0;
3251 }
3252
feb8c6d3
DW		 3253 jbd2_journal_clear_features(sbi->s_journal,
3254 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3255 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3256 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3257 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3258 ret = jbd2_journal_set_features(sbi->s_journal,
3259 compat, 0,
3260 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3261 incompat);
3262 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3263 ret = jbd2_journal_set_features(sbi->s_journal,
3264 compat, 0,
3265 incompat);
3266 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3267 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3268 } else {
feb8c6d3
DW		 3269 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3270 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3271 }
3272
3273 return ret;
3274}
3275
952fc18e
TT		 3276/*
3277 * Note: calculating the overhead so we can be compatible with
3278 * historical BSD practice is quite difficult in the face of
3279 * clusters/bigalloc. This is because multiple metadata blocks from
3280 * different block group can end up in the same allocation cluster.
3281 * Calculating the exact overhead in the face of clustered allocation
3282 * requires either O(all block bitmaps) in memory or O(number of block
3283 * groups**2) in time. We will still calculate the superblock for
3284 * older file systems --- and if we come across with a bigalloc file
3285 * system with zero in s_overhead_clusters the estimate will be close to
3286 * correct especially for very large cluster sizes --- but for newer
3287 * file systems, it's better to calculate this figure once at mkfs
3288 * time, and store it in the superblock. If the superblock value is
3289 * present (even for non-bigalloc file systems), we will use it.
3290 */
3291static int count_overhead(struct super_block *sb, ext4_group_t grp,
3292 char *buf)
3293{
3294 struct ext4_sb_info *sbi = EXT4_SB(sb);
3295 struct ext4_group_desc *gdp;
3296 ext4_fsblk_t first_block, last_block, b;
3297 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3298 int s, j, count = 0;
3299
0548bbb8
TT		 3300 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3301 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3302 sbi->s_itb_per_group + 2);
3303
952fc18e
TT		 3304 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3305 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3306 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3307 for (i = 0; i < ngroups; i++) {
3308 gdp = ext4_get_group_desc(sb, i, NULL);
3309 b = ext4_block_bitmap(sb, gdp);
3310 if (b >= first_block && b <= last_block) {
3311 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3312 count++;
3313 }
3314 b = ext4_inode_bitmap(sb, gdp);
3315 if (b >= first_block && b <= last_block) {
3316 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3317 count++;
3318 }
3319 b = ext4_inode_table(sb, gdp);
3320 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3321 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3322 int c = EXT4_B2C(sbi, b - first_block);
3323 ext4_set_bit(c, buf);
3324 count++;
3325 }
3326 if (i != grp)
3327 continue;
3328 s = 0;
3329 if (ext4_bg_has_super(sb, grp)) {
3330 ext4_set_bit(s++, buf);
3331 count++;
3332 }
3333 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3334 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3335 count++;
3336 }
3337 }
3338 if (!count)
3339 return 0;
3340 return EXT4_CLUSTERS_PER_GROUP(sb) -
3341 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3342}
3343
3344/*
3345 * Compute the overhead and stash it in sbi->s_overhead
3346 */
3347int ext4_calculate_overhead(struct super_block *sb)
3348{
3349 struct ext4_sb_info *sbi = EXT4_SB(sb);
3350 struct ext4_super_block *es = sbi->s_es;
3351 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3352 ext4_fsblk_t overhead = 0;
4fdb5543 3353 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3354
952fc18e
TT		 3355 if (!buf)
3356 return -ENOMEM;
3357
3358 /*
3359 * Compute the overhead (FS structures). This is constant
3360 * for a given filesystem unless the number of block groups
3361 * changes so we cache the previous value until it does.
3362 */
3363
3364 /*
3365 * All of the blocks before first_data_block are overhead
3366 */
3367 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3368
3369 /*
3370 * Add the overhead found in each block group
3371 */
3372 for (i = 0; i < ngroups; i++) {
3373 int blks;
3374
3375 blks = count_overhead(sb, i, buf);
3376 overhead += blks;
3377 if (blks)
3378 memset(buf, 0, PAGE_SIZE);
3379 cond_resched();
3380 }
b003b524
ES		 3381 /* Add the internal journal blocks as well */
3382 if (sbi->s_journal && !sbi->journal_bdev)
810da240 3383 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3384
952fc18e
TT		 3385 sbi->s_overhead = overhead;
3386 smp_wmb();
3387 free_page((unsigned long) buf);
3388 return 0;
3389}
3390
27dd4385 3391
30fac0f7 3392static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3393{
3394 ext4_fsblk_t resv_clusters;
3395
30fac0f7
JK		 3396 /*
3397 * There's no need to reserve anything when we aren't using extents.
3398 * The space estimates are exact, there are no unwritten extents,
3399 * hole punching doesn't need new metadata... This is needed especially
3400 * to keep ext2/3 backward compatibility.
3401 */
3402 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3403 return 0;
27dd4385
LC		 3404 /*
3405 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3406 * This should cover the situations where we can not afford to run
3407 * out of space like for example punch hole, or converting
556615dc 3408 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3409 * allocation would require 1, or 2 blocks, higher numbers are
3410 * very rare.
3411 */
30fac0f7
JK		 3412 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3413 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3414
3415 do_div(resv_clusters, 50);
3416 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3417
3418 return resv_clusters;
3419}
3420
3421
3422static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3423{
3424 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3425 sbi->s_cluster_bits;
3426
3427 if (count >= clusters)
3428 return -EINVAL;
3429
3430 atomic64_set(&sbi->s_resv_clusters, count);
3431 return 0;
3432}
3433
2b2d6d01 3434static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3435{
d4c402d9 3436 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3437 struct buffer_head *bh;
617ba13b
MC		 3438 struct ext4_super_block *es = NULL;
3439 struct ext4_sb_info *sbi;
3440 ext4_fsblk_t block;
3441 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3442 ext4_fsblk_t logical_sb_block;
ac27a0ec 3443 unsigned long offset = 0;
ac27a0ec
DK		 3444 unsigned long journal_devnum = 0;
3445 unsigned long def_mount_opts;
3446 struct inode *root;
9f6200bb 3447 char *cp;
0390131b 3448 const char *descr;
dcc7dae3 3449 int ret = -ENOMEM;
281b5995 3450 int blocksize, clustersize;
4ec11028
TT		 3451 unsigned int db_count;
3452 unsigned int i;
281b5995 3453 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3454 __u64 blocks_count;
07aa2ea1 3455 int err = 0;
b3881f74 3456 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3457 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3458
3459 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3460 if (!sbi)
dcc7dae3 3461 goto out_free_orig;
705895b6
PE		 3462
3463 sbi->s_blockgroup_lock =
3464 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3465 if (!sbi->s_blockgroup_lock) {
3466 kfree(sbi);
dcc7dae3 3467 goto out_free_orig;
705895b6 3468 }
ac27a0ec 3469 sb->s_fs_info = sbi;
2c0544b2 3470 sbi->s_sb = sb;
240799cd 3471 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3472 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3473 if (sb->s_bdev->bd_part)
3474 sbi->s_sectors_written_start =
3475 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3476
9f6200bb
TT		 3477 /* Cleanup superblock name */
3478 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3479 *cp = '!';
3480
07aa2ea1 3481 /* -EINVAL is default */
dcc7dae3 3482 ret = -EINVAL;
617ba13b 3483 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3484 if (!blocksize) {
b31e1552 3485 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3486 goto out_fail;
3487 }
3488
3489 /*
617ba13b 3490 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3491 * block sizes. We need to calculate the offset from buffer start.
3492 */
617ba13b 3493 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3494 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3495 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3496 } else {
70bbb3e0 3497 logical_sb_block = sb_block;
ac27a0ec
DK		 3498 }
3499
a8ac900b 3500 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
b31e1552 3501 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3502 goto out_fail;
3503 }
3504 /*
3505 * Note: s_es must be initialized as soon as possible because
617ba13b 3506 * some ext4 macro-instructions depend on its value
ac27a0ec 3507 */
2716b802 3508 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3509 sbi->s_es = es;
3510 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3511 if (sb->s_magic != EXT4_SUPER_MAGIC)
3512 goto cantfind_ext4;
afc32f7e 3513 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3514
feb0ab32
DW		 3515 /* Warn if metadata_csum and gdt_csum are both set. */
3516 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3517 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3518 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
363307e6 3519 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 3520 "redundant flags; please run fsck.");
3521
d25425f8
DW		 3522 /* Check for a known checksum algorithm */
3523 if (!ext4_verify_csum_type(sb, es)) {
3524 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3525 "unknown checksum algorithm.");
3526 silent = 1;
3527 goto cantfind_ext4;
3528 }
3529
0441984a
DW		 3530 /* Load the checksum driver */
3531 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3532 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3533 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3534 if (IS_ERR(sbi->s_chksum_driver)) {
3535 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3536 ret = PTR_ERR(sbi->s_chksum_driver);
3537 sbi->s_chksum_driver = NULL;
3538 goto failed_mount;
3539 }
3540 }
3541
a9c47317
DW		 3542 /* Check superblock checksum */
3543 if (!ext4_superblock_csum_verify(sb, es)) {
3544 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3545 "invalid superblock checksum. Run e2fsck?");
3546 silent = 1;
3547 goto cantfind_ext4;
3548 }
3549
3550 /* Precompute checksum seed for all metadata */
9aa5d32b 3551 if (ext4_has_metadata_csum(sb))
a9c47317
DW		 3552 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3553 sizeof(es->s_uuid));
3554
ac27a0ec
DK		 3555 /* Set defaults before we parse the mount options */
3556 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3557 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3558 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3559 set_opt(sb, DEBUG);
87f26807 3560 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3561 set_opt(sb, GRPID);
617ba13b 3562 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3563 set_opt(sb, NO_UID32);
ea663336 3564 /* xattr user namespace & acls are now defaulted on */
ea663336 3565 set_opt(sb, XATTR_USER);
03010a33 3566#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3567 set_opt(sb, POSIX_ACL);
2e7842b8 3568#endif
98c1a759
DW		 3569 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3570 if (ext4_has_metadata_csum(sb))
3571 set_opt(sb, JOURNAL_CHECKSUM);
3572
617ba13b 3573 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3574 set_opt(sb, JOURNAL_DATA);
617ba13b 3575 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3576 set_opt(sb, ORDERED_DATA);
617ba13b 3577 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3578 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3579
3580 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3581 set_opt(sb, ERRORS_PANIC);
bb4f397a 3582 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3583 set_opt(sb, ERRORS_CONT);
bb4f397a 3584 else
fd8c37ec 3585 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 3586 /* block_validity enabled by default; disable with noblock_validity */
3587 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3588 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3589 set_opt(sb, DISCARD);
ac27a0ec 3590
08cefc7a
EB		 3591 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3592 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3593 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3594 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3595 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3596
8b67f04a 3597 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3598 set_opt(sb, BARRIER);
ac27a0ec 3599
dd919b98
AK		 3600 /*
3601 * enable delayed allocation by default
3602 * Use -o nodelalloc to turn it off
3603 */
bc0b75f7 3604 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3605 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3606 set_opt(sb, DELALLOC);
dd919b98 3607
51ce6511
LC		 3608 /*
3609 * set default s_li_wait_mult for lazyinit, for the case there is
3610 * no mount option specified.
3611 */
3612 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3613
8b67f04a 3614 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3615 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3616 ext4_msg(sb, KERN_WARNING,
3617 "failed to parse options in superblock: %s",
3618 sbi->s_es->s_mount_opts);
3619 }
5a916be1 3620 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3621 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3622 &journal_ioprio, 0))
ac27a0ec
DK		 3623 goto failed_mount;
3624
56889787
TT		 3625 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3626 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3627 "with data=journal disables delayed "
3628 "allocation and O_DIRECT support!\n");
3629 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3630 ext4_msg(sb, KERN_ERR, "can't mount with "
3631 "both data=journal and delalloc");
3632 goto failed_mount;
3633 }
3634 if (test_opt(sb, DIOREAD_NOLOCK)) {
3635 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3636 "both data=journal and dioread_nolock");
56889787
TT		 3637 goto failed_mount;
3638 }
923ae0ff
RZ		 3639 if (test_opt(sb, DAX)) {
3640 ext4_msg(sb, KERN_ERR, "can't mount with "
3641 "both data=journal and dax");
3642 goto failed_mount;
3643 }
56889787
TT		 3644 if (test_opt(sb, DELALLOC))
3645 clear_opt(sb, DELALLOC);
3646 }
3647
ac27a0ec 3648 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3649 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3650
617ba13b
MC		 3651 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3652 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3653 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3654 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3655 ext4_msg(sb, KERN_WARNING,
3656 "feature flags set on rev 0 fs, "
3657 "running e2fsck is recommended");
469108ff 3658
ed3654eb
TT		 3659 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3660 set_opt2(sb, HURD_COMPAT);
3661 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3662 EXT4_FEATURE_INCOMPAT_64BIT)) {
3663 ext4_msg(sb, KERN_ERR,
3664 "The Hurd can't support 64-bit file systems");
3665 goto failed_mount;
3666 }
3667 }
3668
2035e776
TT		 3669 if (IS_EXT2_SB(sb)) {
3670 if (ext2_feature_set_ok(sb))
3671 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3672 "using the ext4 subsystem");
3673 else {
3674 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3675 "to feature incompatibilities");
3676 goto failed_mount;
3677 }
3678 }
3679
3680 if (IS_EXT3_SB(sb)) {
3681 if (ext3_feature_set_ok(sb))
3682 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3683 "using the ext4 subsystem");
3684 else {
3685 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3686 "to feature incompatibilities");
3687 goto failed_mount;
3688 }
3689 }
3690
ac27a0ec
DK		 3691 /*
3692 * Check feature flags regardless of the revision level, since we
3693 * previously didn't change the revision level when setting the flags,
3694 * so there is a chance incompat flags are set on a rev 0 filesystem.
3695 */
a13fb1a4 3696 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3697 goto failed_mount;
a13fb1a4 3698
261cb20c 3699 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3700 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3701 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3702 ext4_msg(sb, KERN_ERR,
3703 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3704 goto failed_mount;
3705 }
3706
923ae0ff
RZ		 3707 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3708 if (blocksize != PAGE_SIZE) {
3709 ext4_msg(sb, KERN_ERR,
3710 "error: unsupported blocksize for dax");
3711 goto failed_mount;
3712 }
3713 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3714 ext4_msg(sb, KERN_ERR,
3715 "error: device does not support dax");
3716 goto failed_mount;
3717 }
3718 }
3719
6ddb2447
TT		 3720 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3721 es->s_encryption_level) {
3722 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3723 es->s_encryption_level);
3724 goto failed_mount;
3725 }
3726
ac27a0ec 3727 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3728 /* Validate the filesystem blocksize */
3729 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3730 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3731 blocksize);
ac27a0ec
DK		 3732 goto failed_mount;
3733 }
3734
2b2d6d01 3735 brelse(bh);
70bbb3e0
AM		 3736 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3737 offset = do_div(logical_sb_block, blocksize);
a8ac900b 3738 bh = sb_bread_unmovable(sb, logical_sb_block);
ac27a0ec 3739 if (!bh) {
b31e1552
ES		 3740 ext4_msg(sb, KERN_ERR,
3741 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3742 goto failed_mount;
3743 }
2716b802 3744 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3745 sbi->s_es = es;
617ba13b 3746 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3747 ext4_msg(sb, KERN_ERR,
3748 "Magic mismatch, very weird!");
ac27a0ec
DK		 3749 goto failed_mount;
3750 }
3751 }
3752
a13fb1a4
ES		 3753 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3754 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3755 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3756 has_huge_files);
3757 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3758
617ba13b
MC		 3759 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3760 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3761 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3762 } else {
3763 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3764 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3765 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3766 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3767 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3768 ext4_msg(sb, KERN_ERR,
3769 "unsupported inode size: %d",
2b2d6d01 3770 sbi->s_inode_size);
ac27a0ec
DK		 3771 goto failed_mount;
3772 }
ef7f3835
KS		 3773 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3774 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3775 }
0b8e58a1 3776
0d1ee42f
AR		 3777 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3778 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3779 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3780 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3781 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3782 ext4_msg(sb, KERN_ERR,
3783 "unsupported descriptor size %lu",
0d1ee42f
AR		 3784 sbi->s_desc_size);
3785 goto failed_mount;
3786 }
3787 } else
3788 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3789
ac27a0ec 3790 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3791 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3792 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3793 goto cantfind_ext4;
0b8e58a1 3794
617ba13b 3795 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3796 if (sbi->s_inodes_per_block == 0)
617ba13b 3797 goto cantfind_ext4;
ac27a0ec
DK		 3798 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3799 sbi->s_inodes_per_block;
0d1ee42f 3800 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3801 sbi->s_sbh = bh;
3802 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3803 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3804 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3805
2b2d6d01 3806 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3807 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3808 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3809 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3810 i = le32_to_cpu(es->s_flags);
3811 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3812 sbi->s_hash_unsigned = 3;
3813 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3814#ifdef __CHAR_UNSIGNED__
23301410
TT		 3815 if (!(sb->s_flags & MS_RDONLY))
3816 es->s_flags |=
3817 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3818 sbi->s_hash_unsigned = 3;
f99b2589 3819#else
23301410
TT		 3820 if (!(sb->s_flags & MS_RDONLY))
3821 es->s_flags |=
3822 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3823#endif
23301410 3824 }
f99b2589 3825 }
ac27a0ec 3826
281b5995
TT		 3827 /* Handle clustersize */
3828 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3829 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3830 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3831 if (has_bigalloc) {
3832 if (clustersize < blocksize) {
3833 ext4_msg(sb, KERN_ERR,
3834 "cluster size (%d) smaller than "
3835 "block size (%d)", clustersize, blocksize);
3836 goto failed_mount;
3837 }
3838 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3839 le32_to_cpu(es->s_log_block_size);
3840 sbi->s_clusters_per_group =
3841 le32_to_cpu(es->s_clusters_per_group);
3842 if (sbi->s_clusters_per_group > blocksize * 8) {
3843 ext4_msg(sb, KERN_ERR,
3844 "#clusters per group too big: %lu",
3845 sbi->s_clusters_per_group);
3846 goto failed_mount;
3847 }
3848 if (sbi->s_blocks_per_group !=
3849 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3850 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3851 "clusters per group (%lu) inconsistent",
3852 sbi->s_blocks_per_group,
3853 sbi->s_clusters_per_group);
3854 goto failed_mount;
3855 }
3856 } else {
3857 if (clustersize != blocksize) {
3858 ext4_warning(sb, "fragment/cluster size (%d) != "
3859 "block size (%d)", clustersize,
3860 blocksize);
3861 clustersize = blocksize;
3862 }
3863 if (sbi->s_blocks_per_group > blocksize * 8) {
3864 ext4_msg(sb, KERN_ERR,
3865 "#blocks per group too big: %lu",
3866 sbi->s_blocks_per_group);
3867 goto failed_mount;
3868 }
3869 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3870 sbi->s_cluster_bits = 0;
ac27a0ec 3871 }
281b5995
TT		 3872 sbi->s_cluster_ratio = clustersize / blocksize;
3873
ac27a0ec 3874 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3875 ext4_msg(sb, KERN_ERR,
3876 "#inodes per group too big: %lu",
2b2d6d01 3877 sbi->s_inodes_per_group);
ac27a0ec
DK		 3878 goto failed_mount;
3879 }
3880
960fd856
TT		 3881 /* Do we have standard group size of clustersize * 8 blocks ? */
3882 if (sbi->s_blocks_per_group == clustersize << 3)
3883 set_opt2(sb, STD_GROUP_SIZE);
3884
bf43d84b
ES		 3885 /*
3886 * Test whether we have more sectors than will fit in sector_t,
3887 * and whether the max offset is addressable by the page cache.
3888 */
5a9ae68a 3889 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3890 ext4_blocks_count(es));
5a9ae68a 3891 if (err) {
b31e1552 3892 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3893 " too large to mount safely on this system");
ac27a0ec 3894 if (sizeof(sector_t) < 8)
90c699a9 3895 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3896 goto failed_mount;
3897 }
3898
617ba13b
MC		 3899 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3900 goto cantfind_ext4;
e7c95593 3901
0f2ddca6
FTN		 3902 /* check blocks count against device size */
3903 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3904 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3905 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3906 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3907 ext4_blocks_count(es), blocks_count);
3908 goto failed_mount;
3909 }
3910
0b8e58a1
AD		 3911 /*
3912 * It makes no sense for the first data block to be beyond the end
3913 * of the filesystem.
3914 */
3915 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3916 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3917 "block %u is beyond end of filesystem (%llu)",
3918 le32_to_cpu(es->s_first_data_block),
3919 ext4_blocks_count(es));
e7c95593
ES		 3920 goto failed_mount;
3921 }
bd81d8ee
LV		 3922 blocks_count = (ext4_blocks_count(es) -
3923 le32_to_cpu(es->s_first_data_block) +
3924 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3925 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3926 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3927 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3928 "(block count %llu, first data block %u, "
b31e1552 3929 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3930 ext4_blocks_count(es),
3931 le32_to_cpu(es->s_first_data_block),
3932 EXT4_BLOCKS_PER_GROUP(sb));
3933 goto failed_mount;
3934 }
bd81d8ee 3935 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3936 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3937 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3938 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3939 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3940 sbi->s_group_desc = ext4_kvmalloc(db_count *
3941 sizeof(struct buffer_head *),
3942 GFP_KERNEL);
ac27a0ec 3943 if (sbi->s_group_desc == NULL) {
b31e1552 3944 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3945 ret = -ENOMEM;
ac27a0ec
DK		 3946 goto failed_mount;
3947 }
3948
9f6200bb
TT		 3949 if (ext4_proc_root)
3950 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3951
66acdcf4
TT		 3952 if (sbi->s_proc)
3953 proc_create_data("options", S_IRUGO, sbi->s_proc,
3954 &ext4_seq_options_fops, sb);
3955
705895b6 3956 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3957
3958 for (i = 0; i < db_count; i++) {
70bbb3e0 3959 block = descriptor_loc(sb, logical_sb_block, i);
a8ac900b 3960 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
ac27a0ec 3961 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3962 ext4_msg(sb, KERN_ERR,
3963 "can't read group descriptor %d", i);
ac27a0ec
DK		 3964 db_count = i;
3965 goto failed_mount2;
3966 }
3967 }
bfff6873 3968 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3969 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
f9ae9cf5 3970 goto failed_mount2;
ac27a0ec 3971 }
772cb7c8 3972
f9ae9cf5 3973 sbi->s_gdb_count = db_count;
ac27a0ec
DK		 3974 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3975 spin_lock_init(&sbi->s_next_gen_lock);
3976
04ecddb7
JM		 3977 setup_timer(&sbi->s_err_report, print_daily_error_info,
3978 (unsigned long) sb);
04496411 3979
a75ae78f 3980 /* Register extent status tree shrinker */
eb68d0e2 3981 if (ext4_es_register_shrinker(sbi))
ce7e010a 3982 goto failed_mount3;
ce7e010a 3983
c9de560d 3984 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3985 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3986
f9ae9cf5
TT		 3987 /*
3988 * set up enough so that it can read an inode
3989 */
f6e63f90 3990 sb->s_op = &ext4_sops;
617ba13b
MC		 3991 sb->s_export_op = &ext4_export_ops;
3992 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3993#ifdef CONFIG_QUOTA
617ba13b 3994 sb->dq_op = &ext4_quota_operations;
262b4662 3995 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
1fa5efe3 3996 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 3997 else
3998 sb->s_qcop = &ext4_qctl_operations;
96c7e0d9 3999 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
ac27a0ec 4000#endif
f2fa2ffc
AK		 4001 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4002
ac27a0ec 4003 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4004 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 4005
4006 sb->s_root = NULL;
4007
4008 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 4009 EXT4_HAS_INCOMPAT_FEATURE(sb,
4010 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 4011
c5e06d10
JL		 4012 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
4013 !(sb->s_flags & MS_RDONLY))
4014 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4015 goto failed_mount3a;
c5e06d10 4016
ac27a0ec
DK		 4017 /*
4018 * The first inode we look at is the journal inode. Don't try
4019 * root first: it may be modified in the journal!
4020 */
4021 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 4022 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4023 if (ext4_load_journal(sb, es, journal_devnum))
50460fe8 4024 goto failed_mount3a;
0390131b
FM		 4025 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4026 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 4027 ext4_msg(sb, KERN_ERR, "required journal recovery "
4028 "suppressed and not mounted read-only");
744692dc 4029 goto failed_mount_wq;
ac27a0ec 4030 } else {
fd8c37ec 4031 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 4032 sbi->s_journal = NULL;
4033 needs_recovery = 0;
4034 goto no_journal;
ac27a0ec
DK		 4035 }
4036
f32aaf2d 4037 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 4038 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4039 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4040 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4041 goto failed_mount_wq;
eb40a09c
JS		 4042 }
4043
25ed6e8a
DW		 4044 if (!set_journal_csum_feature_set(sb)) {
4045 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4046 "feature set");
4047 goto failed_mount_wq;
d4da6c9c 4048 }
818d276c 4049
ac27a0ec
DK		 4050 /* We have now updated the journal if required, so we can
4051 * validate the data journaling mode. */
4052 switch (test_opt(sb, DATA_FLAGS)) {
4053 case 0:
4054 /* No mode set, assume a default based on the journal
63f57933
AM		 4055 * capabilities: ORDERED_DATA if the journal can
4056 * cope, else JOURNAL_DATA
4057 */
dab291af
MC		 4058 if (jbd2_journal_check_available_features
4059 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4060 set_opt(sb, ORDERED_DATA);
ac27a0ec 4061 else
fd8c37ec 4062 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4063 break;
4064
617ba13b
MC		 4065 case EXT4_MOUNT_ORDERED_DATA:
4066 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4067 if (!jbd2_journal_check_available_features
4068 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4069 ext4_msg(sb, KERN_ERR, "Journal does not support "
4070 "requested data journaling mode");
744692dc 4071 goto failed_mount_wq;
ac27a0ec
DK		 4072 }
4073 default:
4074 break;
4075 }
b3881f74 4076 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4077
18aadd47
BJ		 4078 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4079
ce7e010a 4080no_journal:
9c191f70
M		 4081 if (ext4_mballoc_ready) {
4082 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4083 if (!sbi->s_mb_cache) {
4084 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4085 goto failed_mount_wq;
4086 }
4087 }
4088
1cb767cd
TT		 4089 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
4090 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
4091 (blocksize != PAGE_CACHE_SIZE)) {
4092 ext4_msg(sb, KERN_ERR,
4093 "Unsupported blocksize for fs encryption");
4094 goto failed_mount_wq;
4095 }
4096
4097 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
6ddb2447
TT		 4098 !(sb->s_flags & MS_RDONLY) &&
4099 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4100 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4101 ext4_commit_super(sb, 1);
4102 }
4103
952fc18e
TT		 4104 /*
4105 * Get the # of file system overhead blocks from the
4106 * superblock if present.
4107 */
4108 if (es->s_overhead_clusters)
4109 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4110 else {
07aa2ea1
LC		 4111 err = ext4_calculate_overhead(sb);
4112 if (err)
952fc18e
TT		 4113 goto failed_mount_wq;
4114 }
4115
fd89d5f2
TH		 4116 /*
4117 * The maximum number of concurrent works can be high and
4118 * concurrency isn't really necessary. Limit it to 1.
4119 */
2e8fa54e
JK		 4120 EXT4_SB(sb)->rsv_conversion_wq =
4121 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4122 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4123 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4124 ret = -ENOMEM;
2e8fa54e
JK		 4125 goto failed_mount4;
4126 }
4127
ac27a0ec 4128 /*
dab291af 4129 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4130 * so we can safely mount the rest of the filesystem now.
4131 */
4132
1d1fe1ee
DH		 4133 root = ext4_iget(sb, EXT4_ROOT_INO);
4134 if (IS_ERR(root)) {
b31e1552 4135 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4136 ret = PTR_ERR(root);
32a9bb57 4137 root = NULL;
ac27a0ec
DK		 4138 goto failed_mount4;
4139 }
4140 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4141 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4142 iput(root);
ac27a0ec
DK		 4143 goto failed_mount4;
4144 }
48fde701 4145 sb->s_root = d_make_root(root);
1d1fe1ee 4146 if (!sb->s_root) {
b31e1552 4147 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4148 ret = -ENOMEM;
4149 goto failed_mount4;
4150 }
ac27a0ec 4151
7e84b621
ES		 4152 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4153 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4154
4155 /* determine the minimum size of new large inodes, if present */
4156 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4157 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4158 EXT4_GOOD_OLD_INODE_SIZE;
4159 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4160 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4161 if (sbi->s_want_extra_isize <
4162 le16_to_cpu(es->s_want_extra_isize))
4163 sbi->s_want_extra_isize =
4164 le16_to_cpu(es->s_want_extra_isize);
4165 if (sbi->s_want_extra_isize <
4166 le16_to_cpu(es->s_min_extra_isize))
4167 sbi->s_want_extra_isize =
4168 le16_to_cpu(es->s_min_extra_isize);
4169 }
4170 }
4171 /* Check if enough inode space is available */
4172 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4173 sbi->s_inode_size) {
4174 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4175 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4176 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4177 "available");
ef7f3835
KS		 4178 }
4179
30fac0f7 4180 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4181 if (err) {
4182 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4183 "reserved pool", ext4_calculate_resv_clusters(sb));
f9ae9cf5 4184 goto failed_mount4a;
27dd4385
LC		 4185 }
4186
6fd058f7
TT		 4187 err = ext4_setup_system_zone(sb);
4188 if (err) {
b31e1552 4189 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4190 "zone (%d)", err);
f9ae9cf5
TT		 4191 goto failed_mount4a;
4192 }
4193
4194 ext4_ext_init(sb);
4195 err = ext4_mb_init(sb);
4196 if (err) {
4197 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4198 err);
dcf2d804 4199 goto failed_mount5;
c2774d84
AK		 4200 }
4201
d5e03cbb
TT		 4202 block = ext4_count_free_clusters(sb);
4203 ext4_free_blocks_count_set(sbi->s_es,
4204 EXT4_C2B(sbi, block));
908c7f19
TH		 4205 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4206 GFP_KERNEL);
d5e03cbb
TT		 4207 if (!err) {
4208 unsigned long freei = ext4_count_free_inodes(sb);
4209 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 4210 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4211 GFP_KERNEL);
d5e03cbb
TT		 4212 }
4213 if (!err)
4214 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 4215 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 4216 if (!err)
908c7f19
TH		 4217 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4218 GFP_KERNEL);
d5e03cbb
TT		 4219 if (err) {
4220 ext4_msg(sb, KERN_ERR, "insufficient memory");
4221 goto failed_mount6;
4222 }
4223
4224 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4225 if (!ext4_fill_flex_info(sb)) {
4226 ext4_msg(sb, KERN_ERR,
4227 "unable to initialize "
4228 "flex_bg meta info!");
4229 goto failed_mount6;
4230 }
4231
bfff6873
LC		 4232 err = ext4_register_li_request(sb, first_not_zeroed);
4233 if (err)
dcf2d804 4234 goto failed_mount6;
bfff6873 4235
3197ebdb
TT		 4236 sbi->s_kobj.kset = ext4_kset;
4237 init_completion(&sbi->s_kobj_unregister);
4238 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4239 "%s", sb->s_id);
dcf2d804
TM		 4240 if (err)
4241 goto failed_mount7;
3197ebdb 4242
9b2ff357
JK		 4243#ifdef CONFIG_QUOTA
4244 /* Enable quota usage during mount. */
4245 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4246 !(sb->s_flags & MS_RDONLY)) {
4247 err = ext4_enable_quotas(sb);
4248 if (err)
4249 goto failed_mount8;
4250 }
4251#endif /* CONFIG_QUOTA */
4252
617ba13b
MC		 4253 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4254 ext4_orphan_cleanup(sb, es);
4255 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4256 if (needs_recovery) {
b31e1552 4257 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4258 ext4_mark_recovery_complete(sb, es);
4259 }
4260 if (EXT4_SB(sb)->s_journal) {
4261 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4262 descr = " journalled data mode";
4263 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4264 descr = " ordered data mode";
4265 else
4266 descr = " writeback data mode";
4267 } else
4268 descr = "out journal";
4269
79add3a3
LC		 4270 if (test_opt(sb, DISCARD)) {
4271 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4272 if (!blk_queue_discard(q))
4273 ext4_msg(sb, KERN_WARNING,
4274 "mounting with \"discard\" option, but "
4275 "the device does not support discard");
4276 }
4277
d4c402d9 4278 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4279 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4280 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4281
66e61a9e
TT		 4282 if (es->s_error_count)
4283 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4284
efbed4dc
TT		 4285 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4286 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4287 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4288 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4289
d4c402d9 4290 kfree(orig_data);
ac27a0ec
DK		 4291 return 0;
4292
617ba13b 4293cantfind_ext4:
ac27a0ec 4294 if (!silent)
b31e1552 4295 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4296 goto failed_mount;
4297
72ba7450
TT		 4298#ifdef CONFIG_QUOTA
4299failed_mount8:
4300 kobject_del(&sbi->s_kobj);
4301#endif
dcf2d804
TM		 4302failed_mount7:
4303 ext4_unregister_li_request(sb);
4304failed_mount6:
f9ae9cf5 4305 ext4_mb_release(sb);
d5e03cbb 4306 if (sbi->s_flex_groups)
b93b41d4 4307 kvfree(sbi->s_flex_groups);
d5e03cbb
TT		 4308 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4309 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4310 percpu_counter_destroy(&sbi->s_dirs_counter);
4311 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
00764937 4312failed_mount5:
f9ae9cf5
TT		 4313 ext4_ext_release(sb);
4314 ext4_release_system_zone(sb);
4315failed_mount4a:
94bf608a 4316 dput(sb->s_root);
32a9bb57 4317 sb->s_root = NULL;
94bf608a 4318failed_mount4:
b31e1552 4319 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4320 if (EXT4_SB(sb)->rsv_conversion_wq)
4321 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4322failed_mount_wq:
0390131b
FM		 4323 if (sbi->s_journal) {
4324 jbd2_journal_destroy(sbi->s_journal);
4325 sbi->s_journal = NULL;
4326 }
50460fe8 4327failed_mount3a:
d3922a77 4328 ext4_es_unregister_shrinker(sbi);
eb68d0e2 4329failed_mount3:
9105bb14 4330 del_timer_sync(&sbi->s_err_report);
c5e06d10
JL		 4331 if (sbi->s_mmp_tsk)
4332 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4333failed_mount2:
4334 for (i = 0; i < db_count; i++)
4335 brelse(sbi->s_group_desc[i]);
b93b41d4 4336 kvfree(sbi->s_group_desc);
ac27a0ec 4337failed_mount:
0441984a
DW		 4338 if (sbi->s_chksum_driver)
4339 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4340 if (sbi->s_proc) {
66acdcf4 4341 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4342 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4343 }
ac27a0ec 4344#ifdef CONFIG_QUOTA
a2d4a646 4345 for (i = 0; i < EXT4_MAXQUOTAS; i++)
ac27a0ec
DK		 4346 kfree(sbi->s_qf_names[i]);
4347#endif
617ba13b 4348 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4349 brelse(bh);
4350out_fail:
4351 sb->s_fs_info = NULL;
f6830165 4352 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4353 kfree(sbi);
dcc7dae3 4354out_free_orig:
d4c402d9 4355 kfree(orig_data);
07aa2ea1 4356 return err ? err : ret;
ac27a0ec
DK		 4357}
4358
4359/*
4360 * Setup any per-fs journal parameters now. We'll do this both on
4361 * initial mount, once the journal has been initialised but before we've
4362 * done any recovery; and again on any subsequent remount.
4363 */
617ba13b 4364static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4365{
617ba13b 4366 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4367
30773840
TT		 4368 journal->j_commit_interval = sbi->s_commit_interval;
4369 journal->j_min_batch_time = sbi->s_min_batch_time;
4370 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4371
a931da6a 4372 write_lock(&journal->j_state_lock);
ac27a0ec 4373 if (test_opt(sb, BARRIER))
dab291af 4374 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4375 else
dab291af 4376 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4377 if (test_opt(sb, DATA_ERR_ABORT))
4378 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4379 else
4380 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4381 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4382}
4383
617ba13b 4384static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4385 unsigned int journal_inum)
4386{
4387 struct inode *journal_inode;
4388 journal_t *journal;
4389
0390131b
FM		 4390 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4391
ac27a0ec
DK		 4392 /* First, test for the existence of a valid inode on disk. Bad
4393 * things happen if we iget() an unused inode, as the subsequent
4394 * iput() will try to delete it. */
4395
1d1fe1ee
DH		 4396 journal_inode = ext4_iget(sb, journal_inum);
4397 if (IS_ERR(journal_inode)) {
b31e1552 4398 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4399 return NULL;
4400 }
4401 if (!journal_inode->i_nlink) {
4402 make_bad_inode(journal_inode);
4403 iput(journal_inode);
b31e1552 4404 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4405 return NULL;
4406 }
4407
e5f8eab8 4408 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4409 journal_inode, journal_inode->i_size);
1d1fe1ee 4410 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4411 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4412 iput(journal_inode);
4413 return NULL;
4414 }
4415
dab291af 4416 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4417 if (!journal) {
b31e1552 4418 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4419 iput(journal_inode);
4420 return NULL;
4421 }
4422 journal->j_private = sb;
617ba13b 4423 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4424 return journal;
4425}
4426
617ba13b 4427static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4428 dev_t j_dev)
4429{
2b2d6d01 4430 struct buffer_head *bh;
ac27a0ec 4431 journal_t *journal;
617ba13b
MC		 4432 ext4_fsblk_t start;
4433 ext4_fsblk_t len;
ac27a0ec 4434 int hblock, blocksize;
617ba13b 4435 ext4_fsblk_t sb_block;
ac27a0ec 4436 unsigned long offset;
2b2d6d01 4437 struct ext4_super_block *es;
ac27a0ec
DK		 4438 struct block_device *bdev;
4439
0390131b
FM		 4440 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4441
b31e1552 4442 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4443 if (bdev == NULL)
4444 return NULL;
4445
ac27a0ec 4446 blocksize = sb->s_blocksize;
e1defc4f 4447 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4448 if (blocksize < hblock) {
b31e1552
ES		 4449 ext4_msg(sb, KERN_ERR,
4450 "blocksize too small for journal device");
ac27a0ec
DK		 4451 goto out_bdev;
4452 }
4453
617ba13b
MC		 4454 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4455 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4456 set_blocksize(bdev, blocksize);
4457 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4458 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4459 "external journal");
ac27a0ec
DK		 4460 goto out_bdev;
4461 }
4462
2716b802 4463 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4464 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4465 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4466 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4467 ext4_msg(sb, KERN_ERR, "external journal has "
4468 "bad superblock");
ac27a0ec
DK		 4469 brelse(bh);
4470 goto out_bdev;
4471 }
4472
df4763be
DW		 4473 if ((le32_to_cpu(es->s_feature_ro_compat) &
4474 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4475 es->s_checksum != ext4_superblock_csum(sb, es)) {
4476 ext4_msg(sb, KERN_ERR, "external journal has "
4477 "corrupt superblock");
4478 brelse(bh);
4479 goto out_bdev;
4480 }
4481
617ba13b 4482 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4483 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4484 brelse(bh);
4485 goto out_bdev;
4486 }
4487
bd81d8ee 4488 len = ext4_blocks_count(es);
ac27a0ec
DK		 4489 start = sb_block + 1;
4490 brelse(bh); /* we're done with the superblock */
4491
dab291af 4492 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4493 start, len, blocksize);
4494 if (!journal) {
b31e1552 4495 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4496 goto out_bdev;
4497 }
4498 journal->j_private = sb;
9f203507 4499 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4500 wait_on_buffer(journal->j_sb_buffer);
4501 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4502 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4503 goto out_journal;
4504 }
4505 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4506 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4507 "user (unsupported) - %d",
ac27a0ec
DK		 4508 be32_to_cpu(journal->j_superblock->s_nr_users));
4509 goto out_journal;
4510 }
617ba13b
MC		 4511 EXT4_SB(sb)->journal_bdev = bdev;
4512 ext4_init_journal_params(sb, journal);
ac27a0ec 4513 return journal;
0b8e58a1 4514
ac27a0ec 4515out_journal:
dab291af 4516 jbd2_journal_destroy(journal);
ac27a0ec 4517out_bdev:
617ba13b 4518 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4519 return NULL;
4520}
4521
617ba13b
MC		 4522static int ext4_load_journal(struct super_block *sb,
4523 struct ext4_super_block *es,
ac27a0ec
DK		 4524 unsigned long journal_devnum)
4525{
4526 journal_t *journal;
4527 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4528 dev_t journal_dev;
4529 int err = 0;
4530 int really_read_only;
4531
0390131b
FM		 4532 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4533
ac27a0ec
DK		 4534 if (journal_devnum &&
4535 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4536 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4537 "numbers have changed");
ac27a0ec
DK		 4538 journal_dev = new_decode_dev(journal_devnum);
4539 } else
4540 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4541
4542 really_read_only = bdev_read_only(sb->s_bdev);
4543
4544 /*
4545 * Are we loading a blank journal or performing recovery after a
4546 * crash? For recovery, we need to check in advance whether we
4547 * can get read-write access to the device.
4548 */
617ba13b 4549 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4550 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4551 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4552 "required on readonly filesystem");
ac27a0ec 4553 if (really_read_only) {
b31e1552
ES		 4554 ext4_msg(sb, KERN_ERR, "write access "
4555 "unavailable, cannot proceed");
ac27a0ec
DK		 4556 return -EROFS;
4557 }
b31e1552
ES		 4558 ext4_msg(sb, KERN_INFO, "write access will "
4559 "be enabled during recovery");
ac27a0ec
DK		 4560 }
4561 }
4562
4563 if (journal_inum && journal_dev) {
b31e1552
ES		 4564 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4565 "and inode journals!");
ac27a0ec
DK		 4566 return -EINVAL;
4567 }
4568
4569 if (journal_inum) {
617ba13b 4570 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4571 return -EINVAL;
4572 } else {
617ba13b 4573 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4574 return -EINVAL;
4575 }
4576
90576c0b 4577 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4578 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4579
617ba13b 4580 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4581 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4582 if (!err) {
4583 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4584 if (save)
4585 memcpy(save, ((char *) es) +
4586 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4587 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4588 if (save)
4589 memcpy(((char *) es) + EXT4_S_ERR_START,
4590 save, EXT4_S_ERR_LEN);
4591 kfree(save);
4592 }
ac27a0ec
DK		 4593
4594 if (err) {
b31e1552 4595 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4596 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4597 return err;
4598 }
4599
617ba13b
MC		 4600 EXT4_SB(sb)->s_journal = journal;
4601 ext4_clear_journal_err(sb, es);
ac27a0ec 4602
c41303ce 4603 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4604 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4605 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4606
4607 /* Make sure we flush the recovery flag to disk. */
e2d67052 4608 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4609 }
4610
4611 return 0;
4612}
4613
e2d67052 4614static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4615{
e2d67052 4616 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4617 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4618 int error = 0;
ac27a0ec 4619
08439fec 4620 if (!sbh)
c4be0c1d 4621 return error;
914258bf
TT		 4622 if (buffer_write_io_error(sbh)) {
4623 /*
4624 * Oh, dear. A previous attempt to write the
4625 * superblock failed. This could happen because the
4626 * USB device was yanked out. Or it could happen to
4627 * be a transient write error and maybe the block will
4628 * be remapped. Nothing we can do but to retry the
4629 * write and hope for the best.
4630 */
b31e1552
ES		 4631 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4632 "superblock detected");
914258bf
TT		 4633 clear_buffer_write_io_error(sbh);
4634 set_buffer_uptodate(sbh);
4635 }
71290b36
TT		 4636 /*
4637 * If the file system is mounted read-only, don't update the
4638 * superblock write time. This avoids updating the superblock
4639 * write time when we are mounting the root file system
4640 * read/only but we need to replay the journal; at that point,
4641 * for people who are east of GMT and who make their clock
4642 * tick in localtime for Windows bug-for-bug compatibility,
4643 * the clock is set in the future, and this will cause e2fsck
4644 * to complain and force a full file system check.
4645 */
4646 if (!(sb->s_flags & MS_RDONLY))
4647 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4648 if (sb->s_bdev->bd_part)
4649 es->s_kbytes_written =
4650 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4651 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4652 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4653 else
4654 es->s_kbytes_written =
4655 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
d5e03cbb
TT		 4656 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4657 ext4_free_blocks_count_set(es,
57042651
TT		 4658 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4659 &EXT4_SB(sb)->s_freeclusters_counter)));
d5e03cbb
TT		 4660 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4661 es->s_free_inodes_count =
4662 cpu_to_le32(percpu_counter_sum_positive(
ce7e010a 4663 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4664 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4665 ext4_superblock_csum_set(sb);
ac27a0ec 4666 mark_buffer_dirty(sbh);
914258bf 4667 if (sync) {
c4be0c1d
TS		 4668 error = sync_dirty_buffer(sbh);
4669 if (error)
4670 return error;
4671
4672 error = buffer_write_io_error(sbh);
4673 if (error) {
b31e1552
ES		 4674 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4675 "superblock");
914258bf
TT		 4676 clear_buffer_write_io_error(sbh);
4677 set_buffer_uptodate(sbh);
4678 }
4679 }
c4be0c1d 4680 return error;
ac27a0ec
DK		 4681}
4682
ac27a0ec
DK		 4683/*
4684 * Have we just finished recovery? If so, and if we are mounting (or
4685 * remounting) the filesystem readonly, then we will end up with a
4686 * consistent fs on disk. Record that fact.
4687 */
2b2d6d01
TT		 4688static void ext4_mark_recovery_complete(struct super_block *sb,
4689 struct ext4_super_block *es)
ac27a0ec 4690{
617ba13b 4691 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4692
0390131b
FM		 4693 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4694 BUG_ON(journal != NULL);
4695 return;
4696 }
dab291af 4697 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4698 if (jbd2_journal_flush(journal) < 0)
4699 goto out;
4700
617ba13b 4701 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4702 sb->s_flags & MS_RDONLY) {
617ba13b 4703 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4704 ext4_commit_super(sb, 1);
ac27a0ec 4705 }
7ffe1ea8
HK		 4706
4707out:
dab291af 4708 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4709}
4710
4711/*
4712 * If we are mounting (or read-write remounting) a filesystem whose journal
4713 * has recorded an error from a previous lifetime, move that error to the
4714 * main filesystem now.
4715 */
2b2d6d01
TT		 4716static void ext4_clear_journal_err(struct super_block *sb,
4717 struct ext4_super_block *es)
ac27a0ec
DK		 4718{
4719 journal_t *journal;
4720 int j_errno;
4721 const char *errstr;
4722
0390131b
FM		 4723 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4724
617ba13b 4725 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4726
4727 /*
4728 * Now check for any error status which may have been recorded in the
617ba13b 4729 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4730 */
4731
dab291af 4732 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4733 if (j_errno) {
4734 char nbuf[16];
4735
617ba13b 4736 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4737 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4738 "from previous mount: %s", errstr);
12062ddd 4739 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4740
617ba13b
MC		 4741 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4742 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4743 ext4_commit_super(sb, 1);
ac27a0ec 4744
dab291af 4745 jbd2_journal_clear_err(journal);
d796c52e 4746 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4747 }
4748}
4749
4750/*
4751 * Force the running and committing transactions to commit,
4752 * and wait on the commit.
4753 */
617ba13b 4754int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4755{
4756 journal_t *journal;
ac27a0ec
DK		 4757
4758 if (sb->s_flags & MS_RDONLY)
4759 return 0;
4760
617ba13b 4761 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4762 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4763}
4764
617ba13b 4765static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4766{
14ce0cb4 4767 int ret = 0;
9eddacf9 4768 tid_t target;
06a407f1 4769 bool needs_barrier = false;
8d5d02e6 4770 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4771
9bffad1e 4772 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4773 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4774 /*
4775 * Writeback quota in non-journalled quota case - journalled quota has
4776 * no dirty dquots
4777 */
4778 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4779 /*
4780 * Data writeback is possible w/o journal transaction, so barrier must
4781 * being sent at the end of the function. But we can skip it if
4782 * transaction_commit will do it for us.
4783 */
bda32530
TT		 4784 if (sbi->s_journal) {
4785 target = jbd2_get_latest_transaction(sbi->s_journal);
4786 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4787 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4788 needs_barrier = true;
4789
4790 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4791 if (wait)
4792 ret = jbd2_log_wait_commit(sbi->s_journal,
4793 target);
4794 }
4795 } else if (wait && test_opt(sb, BARRIER))
06a407f1 4796 needs_barrier = true;
06a407f1
DM		 4797 if (needs_barrier) {
4798 int err;
4799 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4800 if (!ret)
4801 ret = err;
0390131b 4802 }
06a407f1
DM		 4803
4804 return ret;
4805}
4806
ac27a0ec
DK		 4807/*
4808 * LVM calls this function before a (read-only) snapshot is created. This
4809 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4810 *
4811 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4812 * state independently. It relies on upper layer to stop all data & metadata
4813 * modifications.
ac27a0ec 4814 */
c4be0c1d 4815static int ext4_freeze(struct super_block *sb)
ac27a0ec 4816{
c4be0c1d
TS		 4817 int error = 0;
4818 journal_t *journal;
ac27a0ec 4819
9ca92389
TT		 4820 if (sb->s_flags & MS_RDONLY)
4821 return 0;
ac27a0ec 4822
9ca92389 4823 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4824
bb044576
TT		 4825 if (journal) {
4826 /* Now we set up the journal barrier. */
4827 jbd2_journal_lock_updates(journal);
ac27a0ec 4828
bb044576
TT		 4829 /*
4830 * Don't clear the needs_recovery flag if we failed to
4831 * flush the journal.
4832 */
4833 error = jbd2_journal_flush(journal);
4834 if (error < 0)
4835 goto out;
4836 }
9ca92389
TT		 4837
4838 /* Journal blocked and flushed, clear needs_recovery flag. */
4839 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4840 error = ext4_commit_super(sb, 1);
6b0310fb 4841out:
bb044576
TT		 4842 if (journal)
4843 /* we rely on upper layer to stop further updates */
4844 jbd2_journal_unlock_updates(journal);
6b0310fb 4845 return error;
ac27a0ec
DK		 4846}
4847
4848/*
4849 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4850 * flag here, even though the filesystem is not technically dirty yet.
4851 */
c4be0c1d 4852static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4853{
9ca92389
TT		 4854 if (sb->s_flags & MS_RDONLY)
4855 return 0;
4856
9ca92389
TT		 4857 /* Reset the needs_recovery flag before the fs is unlocked. */
4858 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4859 ext4_commit_super(sb, 1);
c4be0c1d 4860 return 0;
ac27a0ec
DK		 4861}
4862
673c6100
TT		 4863/*
4864 * Structure to save mount options for ext4_remount's benefit
4865 */
4866struct ext4_mount_options {
4867 unsigned long s_mount_opt;
a2595b8a 4868 unsigned long s_mount_opt2;
08cefc7a
EB		 4869 kuid_t s_resuid;
4870 kgid_t s_resgid;
673c6100
TT		 4871 unsigned long s_commit_interval;
4872 u32 s_min_batch_time, s_max_batch_time;
4873#ifdef CONFIG_QUOTA
4874 int s_jquota_fmt;
a2d4a646 4875 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 4876#endif
4877};
4878
2b2d6d01 4879static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4880{
2b2d6d01 4881 struct ext4_super_block *es;
617ba13b 4882 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4883 unsigned long old_sb_flags;
617ba13b 4884 struct ext4_mount_options old_opts;
c79d967d 4885 int enable_quota = 0;
8a266467 4886 ext4_group_t g;
b3881f74 4887 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4888 int err = 0;
ac27a0ec 4889#ifdef CONFIG_QUOTA
03dafb5f 4890 int i, j;
ac27a0ec 4891#endif
d4c402d9 4892 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4893
4894 /* Store the original options */
4895 old_sb_flags = sb->s_flags;
4896 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4897 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4898 old_opts.s_resuid = sbi->s_resuid;
4899 old_opts.s_resgid = sbi->s_resgid;
4900 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4901 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4902 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4903#ifdef CONFIG_QUOTA
4904 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 4905 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f
CG		 4906 if (sbi->s_qf_names[i]) {
4907 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4908 GFP_KERNEL);
4909 if (!old_opts.s_qf_names[i]) {
4910 for (j = 0; j < i; j++)
4911 kfree(old_opts.s_qf_names[j]);
3e36a163 4912 kfree(orig_data);
03dafb5f
CG		 4913 return -ENOMEM;
4914 }
4915 } else
4916 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4917#endif
b3881f74
TT		 4918 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4919 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 4920
661aa520 4921 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4922 err = -EINVAL;
4923 goto restore_opts;
4924 }
4925
6b992ff2 4926 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 4927 test_opt(sb, JOURNAL_CHECKSUM)) {
4928 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 4929 "during remount not supported; ignoring");
4930 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 4931 }
4932
6ae6514b
PS		 4933 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4934 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4935 ext4_msg(sb, KERN_ERR, "can't mount with "
4936 "both data=journal and delalloc");
4937 err = -EINVAL;
4938 goto restore_opts;
4939 }
4940 if (test_opt(sb, DIOREAD_NOLOCK)) {
4941 ext4_msg(sb, KERN_ERR, "can't mount with "
4942 "both data=journal and dioread_nolock");
4943 err = -EINVAL;
4944 goto restore_opts;
4945 }
923ae0ff
RZ		 4946 if (test_opt(sb, DAX)) {
4947 ext4_msg(sb, KERN_ERR, "can't mount with "
4948 "both data=journal and dax");
4949 err = -EINVAL;
4950 goto restore_opts;
4951 }
4952 }
4953
4954 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4955 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4956 "dax flag with busy inodes while remounting");
4957 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
6ae6514b
PS		 4958 }
4959
4ab2f15b 4960 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4961 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4962
4963 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4964 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4965
4966 es = sbi->s_es;
4967
b3881f74 4968 if (sbi->s_journal) {
0390131b 4969 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4970 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4971 }
ac27a0ec 4972
661aa520 4973 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4974 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4975 err = -EROFS;
4976 goto restore_opts;
4977 }
4978
4979 if (*flags & MS_RDONLY) {
38c03b34
TT		 4980 err = sync_filesystem(sb);
4981 if (err < 0)
4982 goto restore_opts;
0f0dd62f
CH		 4983 err = dquot_suspend(sb, -1);
4984 if (err < 0)
c79d967d 4985 goto restore_opts;
c79d967d 4986
ac27a0ec
DK		 4987 /*
4988 * First of all, the unconditional stuff we have to do
4989 * to disable replay of the journal when we next remount
4990 */
4991 sb->s_flags |= MS_RDONLY;
4992
4993 /*
4994 * OK, test if we are remounting a valid rw partition
4995 * readonly, and if so set the rdonly flag and then
4996 * mark the partition as valid again.
4997 */
617ba13b
MC		 4998 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4999 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5000 es->s_state = cpu_to_le16(sbi->s_mount_state);
5001
a63c9eb2 5002 if (sbi->s_journal)
0390131b 5003 ext4_mark_recovery_complete(sb, es);
ac27a0ec 5004 } else {
a13fb1a4 5005 /* Make sure we can mount this feature set readwrite */
2cb5cc8b
DW		 5006 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5007 EXT4_FEATURE_RO_COMPAT_READONLY) ||
5008 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5009 err = -EROFS;
5010 goto restore_opts;
5011 }
8a266467
TT		 5012 /*
5013 * Make sure the group descriptor checksums
0b8e58a1 5014 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5015 */
5016 for (g = 0; g < sbi->s_groups_count; g++) {
5017 struct ext4_group_desc *gdp =
5018 ext4_get_group_desc(sb, g, NULL);
5019
feb0ab32 5020 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5021 ext4_msg(sb, KERN_ERR,
5022 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 5023 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
5024 le16_to_cpu(gdp->bg_checksum));
5025 err = -EINVAL;
5026 goto restore_opts;
5027 }
5028 }
5029
ead6596b
ES		 5030 /*
5031 * If we have an unprocessed orphan list hanging
5032 * around from a previously readonly bdev mount,
5033 * require a full umount/remount for now.
5034 */
5035 if (es->s_last_orphan) {
b31e1552 5036 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5037 "remount RDWR because of unprocessed "
5038 "orphan inode list. Please "
b31e1552 5039 "umount/remount instead");
ead6596b
ES		 5040 err = -EINVAL;
5041 goto restore_opts;
5042 }
5043
ac27a0ec
DK		 5044 /*
5045 * Mounting a RDONLY partition read-write, so reread
5046 * and store the current valid flag. (It may have
5047 * been changed by e2fsck since we originally mounted
5048 * the partition.)
5049 */
0390131b
FM		 5050 if (sbi->s_journal)
5051 ext4_clear_journal_err(sb, es);
ac27a0ec 5052 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 5053 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 5054 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 5055 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5056 EXT4_FEATURE_INCOMPAT_MMP))
5057 if (ext4_multi_mount_protect(sb,
5058 le64_to_cpu(es->s_mmp_block))) {
5059 err = -EROFS;
5060 goto restore_opts;
5061 }
c79d967d 5062 enable_quota = 1;
ac27a0ec
DK		 5063 }
5064 }
bfff6873
LC		 5065
5066 /*
5067 * Reinitialize lazy itable initialization thread based on
5068 * current settings
5069 */
5070 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5071 ext4_unregister_li_request(sb);
5072 else {
5073 ext4_group_t first_not_zeroed;
5074 first_not_zeroed = ext4_has_uninit_itable(sb);
5075 ext4_register_li_request(sb, first_not_zeroed);
5076 }
5077
6fd058f7 5078 ext4_setup_system_zone(sb);
d096ad0f 5079 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 5080 ext4_commit_super(sb, 1);
0390131b 5081
ac27a0ec
DK		 5082#ifdef CONFIG_QUOTA
5083 /* Release old quota file names */
a2d4a646 5084 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5085 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 5086 if (enable_quota) {
5087 if (sb_any_quota_suspended(sb))
5088 dquot_resume(sb, -1);
5089 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5090 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5091 err = ext4_enable_quotas(sb);
07724f98 5092 if (err)
7c319d32 5093 goto restore_opts;
7c319d32
AK		 5094 }
5095 }
ac27a0ec 5096#endif
d4c402d9 5097
a26f4992 5098 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
d4c402d9
CW		 5099 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5100 kfree(orig_data);
ac27a0ec 5101 return 0;
0b8e58a1 5102
ac27a0ec
DK		 5103restore_opts:
5104 sb->s_flags = old_sb_flags;
5105 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 5106 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 5107 sbi->s_resuid = old_opts.s_resuid;
5108 sbi->s_resgid = old_opts.s_resgid;
5109 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5110 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5111 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5112#ifdef CONFIG_QUOTA
5113 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 5114 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
03dafb5f 5115 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5116 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5117 }
5118#endif
d4c402d9 5119 kfree(orig_data);
ac27a0ec
DK		 5120 return err;
5121}
5122
2b2d6d01 5123static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5124{
5125 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5126 struct ext4_sb_info *sbi = EXT4_SB(sb);
5127 struct ext4_super_block *es = sbi->s_es;
27dd4385 5128 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5129 u64 fsid;
d02a9391 5130 s64 bfree;
27dd4385 5131 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5132
952fc18e
TT		 5133 if (!test_opt(sb, MINIX_DF))
5134 overhead = sbi->s_overhead;
ac27a0ec 5135
617ba13b 5136 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5137 buf->f_bsize = sb->s_blocksize;
b72f78cb 5138 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5139 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5140 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5141 /* prevent underflow in case that few free space is available */
57042651 5142 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5143 buf->f_bavail = buf->f_bfree -
5144 (ext4_r_blocks_count(es) + resv_blocks);
5145 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5146 buf->f_bavail = 0;
5147 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5148 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5149 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5150 fsid = le64_to_cpup((void *)es->s_uuid) ^
5151 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5152 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5153 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5154
ac27a0ec
DK		 5155 return 0;
5156}
5157
0b8e58a1
AD		 5158/* Helper function for writing quotas on sync - we need to start transaction
5159 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5160 * Process 1 Process 2
617ba13b 5161 * ext4_create() quota_sync()
a269eb18 5162 * jbd2_journal_start() write_dquot()
871a2931 5163 * dquot_initialize() down(dqio_mutex)
dab291af 5164 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5165 *
5166 */
5167
5168#ifdef CONFIG_QUOTA
5169
5170static inline struct inode *dquot_to_inode(struct dquot *dquot)
5171{
4c376dca 5172 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5173}
5174
617ba13b 5175static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5176{
5177 int ret, err;
5178 handle_t *handle;
5179 struct inode *inode;
5180
5181 inode = dquot_to_inode(dquot);
9924a92a 5182 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5183 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5184 if (IS_ERR(handle))
5185 return PTR_ERR(handle);
5186 ret = dquot_commit(dquot);
617ba13b 5187 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5188 if (!ret)
5189 ret = err;
5190 return ret;
5191}
5192
617ba13b 5193static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5194{
5195 int ret, err;
5196 handle_t *handle;
5197
9924a92a 5198 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5199 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5200 if (IS_ERR(handle))
5201 return PTR_ERR(handle);
5202 ret = dquot_acquire(dquot);
617ba13b 5203 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5204 if (!ret)
5205 ret = err;
5206 return ret;
5207}
5208
617ba13b 5209static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5210{
5211 int ret, err;
5212 handle_t *handle;
5213
9924a92a 5214 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5215 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5216 if (IS_ERR(handle)) {
5217 /* Release dquot anyway to avoid endless cycle in dqput() */
5218 dquot_release(dquot);
ac27a0ec 5219 return PTR_ERR(handle);
9c3013e9 5220 }
ac27a0ec 5221 ret = dquot_release(dquot);
617ba13b 5222 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5223 if (!ret)
5224 ret = err;
5225 return ret;
5226}
5227
617ba13b 5228static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5229{
262b4662
JK		 5230 struct super_block *sb = dquot->dq_sb;
5231 struct ext4_sb_info *sbi = EXT4_SB(sb);
5232
2c8be6b2 5233 /* Are we journaling quotas? */
262b4662
JK		 5234 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5235 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5236 dquot_mark_dquot_dirty(dquot);
617ba13b 5237 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5238 } else {
5239 return dquot_mark_dquot_dirty(dquot);
5240 }
5241}
5242
617ba13b 5243static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5244{
5245 int ret, err;
5246 handle_t *handle;
5247
5248 /* Data block + inode block */
2b0143b5 5249 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5250 if (IS_ERR(handle))
5251 return PTR_ERR(handle);
5252 ret = dquot_commit_info(sb, type);
617ba13b 5253 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5254 if (!ret)
5255 ret = err;
5256 return ret;
5257}
5258
5259/*
5260 * Turn on quotas during mount time - we need to find
5261 * the quota file and such...
5262 */
617ba13b 5263static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5264{
287a8095
CH		 5265 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5266 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5267}
5268
5269/*
5270 * Standard function to be called on quota_on
5271 */
617ba13b 5272static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5273 struct path *path)
ac27a0ec
DK		 5274{
5275 int err;
ac27a0ec
DK		 5276
5277 if (!test_opt(sb, QUOTA))
5278 return -EINVAL;
0623543b 5279
ac27a0ec 5280 /* Quotafile not on the same filesystem? */
d8c9584e 5281 if (path->dentry->d_sb != sb)
ac27a0ec 5282 return -EXDEV;
0623543b
JK		 5283 /* Journaling quota? */
5284 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5285 /* Quotafile not in fs root? */
f00c9e44 5286 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5287 ext4_msg(sb, KERN_WARNING,
5288 "Quota file not on filesystem root. "
5289 "Journaled quota will not work");
2b2d6d01 5290 }
0623543b
JK		 5291
5292 /*
5293 * When we journal data on quota file, we have to flush journal to see
5294 * all updates to the file when we bypass pagecache...
5295 */
0390131b 5296 if (EXT4_SB(sb)->s_journal &&
2b0143b5 5297 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 5298 /*
5299 * We don't need to lock updates but journal_flush() could
5300 * otherwise be livelocked...
5301 */
5302 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5303 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5304 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5305 if (err)
7ffe1ea8 5306 return err;
0623543b
JK		 5307 }
5308
f00c9e44 5309 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5310}
5311
7c319d32
AK		 5312static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5313 unsigned int flags)
5314{
5315 int err;
5316 struct inode *qf_inode;
a2d4a646 5317 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5318 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5319 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5320 };
5321
5322 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5323
5324 if (!qf_inums[type])
5325 return -EPERM;
5326
5327 qf_inode = ext4_iget(sb, qf_inums[type]);
5328 if (IS_ERR(qf_inode)) {
5329 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5330 return PTR_ERR(qf_inode);
5331 }
5332
bcb13850
JK		 5333 /* Don't account quota for quota files to avoid recursion */
5334 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5335 err = dquot_enable(qf_inode, type, format_id, flags);
5336 iput(qf_inode);
5337
5338 return err;
5339}
5340
5341/* Enable usage tracking for all quota types. */
5342static int ext4_enable_quotas(struct super_block *sb)
5343{
5344 int type, err = 0;
a2d4a646 5345 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32
AK		 5346 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5347 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5348 };
5349
5350 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
a2d4a646 5351 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 5352 if (qf_inums[type]) {
5353 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5354 DQUOT_USAGE_ENABLED);
5355 if (err) {
5356 ext4_warning(sb,
72ba7450
TT		 5357 "Failed to enable quota tracking "
5358 "(type=%d, err=%d). Please run "
5359 "e2fsck to fix.", type, err);
7c319d32
AK		 5360 return err;
5361 }
5362 }
5363 }
5364 return 0;
5365}
5366
ca0e05e4
DM		 5367static int ext4_quota_off(struct super_block *sb, int type)
5368{
21f97697
JK		 5369 struct inode *inode = sb_dqopt(sb)->files[type];
5370 handle_t *handle;
5371
87009d86
DM		 5372 /* Force all delayed allocation blocks to be allocated.
5373 * Caller already holds s_umount sem */
5374 if (test_opt(sb, DELALLOC))
ca0e05e4 5375 sync_filesystem(sb);
ca0e05e4 5376
0b268590
AG		 5377 if (!inode)
5378 goto out;
5379
21f97697
JK		 5380 /* Update modification times of quota files when userspace can
5381 * start looking at them */
9924a92a 5382 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5383 if (IS_ERR(handle))
5384 goto out;
5385 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5386 ext4_mark_inode_dirty(handle, inode);
5387 ext4_journal_stop(handle);
5388
5389out:
ca0e05e4
DM		 5390 return dquot_quota_off(sb, type);
5391}
5392
ac27a0ec
DK		 5393/* Read data from quotafile - avoid pagecache and such because we cannot afford
5394 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5395 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5396 * we don't have to be afraid of races */
617ba13b 5397static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5398 size_t len, loff_t off)
5399{
5400 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5401 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5402 int offset = off & (sb->s_blocksize - 1);
5403 int tocopy;
5404 size_t toread;
5405 struct buffer_head *bh;
5406 loff_t i_size = i_size_read(inode);
5407
5408 if (off > i_size)
5409 return 0;
5410 if (off+len > i_size)
5411 len = i_size-off;
5412 toread = len;
5413 while (toread > 0) {
5414 tocopy = sb->s_blocksize - offset < toread ?
5415 sb->s_blocksize - offset : toread;
1c215028
TT		 5416 bh = ext4_bread(NULL, inode, blk, 0);
5417 if (IS_ERR(bh))
5418 return PTR_ERR(bh);
ac27a0ec
DK		 5419 if (!bh) /* A hole? */
5420 memset(data, 0, tocopy);
5421 else
5422 memcpy(data, bh->b_data+offset, tocopy);
5423 brelse(bh);
5424 offset = 0;
5425 toread -= tocopy;
5426 data += tocopy;
5427 blk++;
5428 }
5429 return len;
5430}
5431
5432/* Write to quotafile (we know the transaction is already started and has
5433 * enough credits) */
617ba13b 5434static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5435 const char *data, size_t len, loff_t off)
5436{
5437 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5438 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
1c215028 5439 int err, offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5440 struct buffer_head *bh;
5441 handle_t *handle = journal_current_handle();
5442
0390131b 5443 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5444 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5445 " cancelled because transaction is not started",
9c3013e9
JK		 5446 (unsigned long long)off, (unsigned long long)len);
5447 return -EIO;
5448 }
67eeb568
DM		 5449 /*
5450 * Since we account only one data block in transaction credits,
5451 * then it is impossible to cross a block boundary.
5452 */
5453 if (sb->s_blocksize - offset < len) {
5454 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5455 " cancelled because not block aligned",
5456 (unsigned long long)off, (unsigned long long)len);
5457 return -EIO;
5458 }
5459
1c215028
TT		 5460 bh = ext4_bread(handle, inode, blk, 1);
5461 if (IS_ERR(bh))
5462 return PTR_ERR(bh);
67eeb568
DM		 5463 if (!bh)
5464 goto out;
5d601255 5465 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5466 err = ext4_journal_get_write_access(handle, bh);
5467 if (err) {
5468 brelse(bh);
1c215028 5469 return err;
ac27a0ec 5470 }
67eeb568
DM		 5471 lock_buffer(bh);
5472 memcpy(bh->b_data+offset, data, len);
5473 flush_dcache_page(bh->b_page);
5474 unlock_buffer(bh);
62d2b5f2 5475 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5476 brelse(bh);
ac27a0ec 5477out:
67eeb568
DM		 5478 if (inode->i_size < off + len) {
5479 i_size_write(inode, off + len);
617ba13b 5480 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5481 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5482 }
67eeb568 5483 return len;
ac27a0ec
DK		 5484}
5485
5486#endif
5487
152a0836
AV		 5488static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5489 const char *dev_name, void *data)
ac27a0ec 5490{
152a0836 5491 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5492}
5493
37f328eb 5494#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5495static inline void register_as_ext2(void)
5496{
5497 int err = register_filesystem(&ext2_fs_type);
5498 if (err)
5499 printk(KERN_WARNING
5500 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5501}
5502
5503static inline void unregister_as_ext2(void)
5504{
5505 unregister_filesystem(&ext2_fs_type);
5506}
2035e776
TT		 5507
5508static inline int ext2_feature_set_ok(struct super_block *sb)
5509{
5510 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5511 return 0;
5512 if (sb->s_flags & MS_RDONLY)
5513 return 1;
5514 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5515 return 0;
5516 return 1;
5517}
24b58424
TT		 5518#else
5519static inline void register_as_ext2(void) { }
5520static inline void unregister_as_ext2(void) { }
2035e776 5521static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5522#endif
5523
37f328eb 5524#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5525static inline void register_as_ext3(void)
5526{
5527 int err = register_filesystem(&ext3_fs_type);
5528 if (err)
5529 printk(KERN_WARNING
5530 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5531}
5532
5533static inline void unregister_as_ext3(void)
5534{
5535 unregister_filesystem(&ext3_fs_type);
5536}
2035e776
TT		 5537
5538static inline int ext3_feature_set_ok(struct super_block *sb)
5539{
5540 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5541 return 0;
5542 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5543 return 0;
5544 if (sb->s_flags & MS_RDONLY)
5545 return 1;
5546 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5547 return 0;
5548 return 1;
5549}
24b58424
TT		 5550#else
5551static inline void register_as_ext3(void) { }
5552static inline void unregister_as_ext3(void) { }
2035e776 5553static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5554#endif
5555
03010a33
TT		 5556static struct file_system_type ext4_fs_type = {
5557 .owner = THIS_MODULE,
5558 .name = "ext4",
152a0836 5559 .mount = ext4_mount,
03010a33
TT		 5560 .kill_sb = kill_block_super,
5561 .fs_flags = FS_REQUIRES_DEV,
5562};
7f78e035 5563MODULE_ALIAS_FS("ext4");
03010a33 5564
8f021222 5565static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5566{
5567 struct ext4_features *ef;
5568 int ret = -ENOMEM;
5569
5570 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5571 if (!ef)
5572 goto out;
5573
5574 ef->f_kobj.kset = ext4_kset;
5575 init_completion(&ef->f_kobj_unregister);
5576 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5577 "features");
5578 if (ret) {
5579 kfree(ef);
5580 goto out;
5581 }
5582
5583 ext4_feat = ef;
5584 ret = 0;
5585out:
5586 return ret;
5587}
5588
8f021222
LC		 5589static void ext4_exit_feat_adverts(void)
5590{
5591 kobject_put(&ext4_feat->f_kobj);
5592 wait_for_completion(&ext4_feat->f_kobj_unregister);
5593 kfree(ext4_feat);
5594}
5595
e9e3bcec
ES		 5596/* Shared across all ext4 file systems */
5597wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5598struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5599
5dabfc78 5600static int __init ext4_init_fs(void)
ac27a0ec 5601{
e9e3bcec 5602 int i, err;
c9de560d 5603
07c0c5d8
AV		 5604 ext4_li_info = NULL;
5605 mutex_init(&ext4_li_mtx);
5606
9a4c8019 5607 /* Build-time check for flags consistency */
12e9b892 5608 ext4_check_flag_values();
e9e3bcec
ES		 5609
5610 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5611 mutex_init(&ext4__aio_mutex[i]);
5612 init_waitqueue_head(&ext4__ioend_wq[i]);
5613 }
5614
51865fda 5615 err = ext4_init_es();
6fd058f7
TT		 5616 if (err)
5617 return err;
51865fda
ZL		 5618
5619 err = ext4_init_pageio();
5620 if (err)
5621 goto out7;
5622
5dabfc78 5623 err = ext4_init_system_zone();
bd2d0210 5624 if (err)
d44651d0 5625 goto out6;
3197ebdb 5626 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5627 if (!ext4_kset) {
5628 err = -ENOMEM;
dd68314c 5629 goto out5;
0e376b1e 5630 }
d44651d0 5631 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5632
5633 err = ext4_init_feat_adverts();
dd68314c
TT		 5634 if (err)
5635 goto out4;
857ac889 5636
5dabfc78 5637 err = ext4_init_mballoc();
c9de560d
AT		 5638 if (err)
5639 goto out2;
9c191f70
M		 5640 else
5641 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5642 err = init_inodecache();
5643 if (err)
5644 goto out1;
24b58424 5645 register_as_ext3();
2035e776 5646 register_as_ext2();
03010a33 5647 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5648 if (err)
5649 goto out;
bfff6873 5650
ac27a0ec
DK		 5651 return 0;
5652out:
24b58424
TT		 5653 unregister_as_ext2();
5654 unregister_as_ext3();
ac27a0ec
DK		 5655 destroy_inodecache();
5656out1:
9c191f70 5657 ext4_mballoc_ready = 0;
5dabfc78 5658 ext4_exit_mballoc();
9c191f70 5659out2:
8f021222 5660 ext4_exit_feat_adverts();
dd68314c 5661out4:
d44651d0
FJ		 5662 if (ext4_proc_root)
5663 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5664 kset_unregister(ext4_kset);
d44651d0 5665out5:
5dabfc78 5666 ext4_exit_system_zone();
d44651d0 5667out6:
5dabfc78 5668 ext4_exit_pageio();
51865fda
ZL		 5669out7:
5670 ext4_exit_es();
5671
ac27a0ec
DK		 5672 return err;
5673}
5674
5dabfc78 5675static void __exit ext4_exit_fs(void)
ac27a0ec 5676{
e298e73b 5677 ext4_exit_crypto();
bfff6873 5678 ext4_destroy_lazyinit_thread();
24b58424
TT		 5679 unregister_as_ext2();
5680 unregister_as_ext3();
03010a33 5681 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5682 destroy_inodecache();
5dabfc78 5683 ext4_exit_mballoc();
8f021222 5684 ext4_exit_feat_adverts();
9f6200bb 5685 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5686 kset_unregister(ext4_kset);
5dabfc78
TT		 5687 ext4_exit_system_zone();
5688 ext4_exit_pageio();
dd12ed14 5689 ext4_exit_es();
ac27a0ec
DK		 5690}
5691
5692MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5693MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5694MODULE_LICENSE("GPL");
5dabfc78
TT		 5695module_init(ext4_init_fs)
5696module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository